From c9cbd6d832d39142e203435a84acfe0d3972bf8f Mon Sep 17 00:00:00 2001 From: Hamza Date: Mon, 4 Aug 2025 11:25:42 +0300 Subject: [PATCH 1/4] Update personal site configuration, remove unused announcements, and enhance project descriptions. Revise bio to reflect new roles and achievements, including founding HoverNotes and publishing research. Update skills and project details for clarity and relevance. --- Gemfile | 1 - _config.yml | 6 +- _data/cv.yml | 77 ++++++++++++++++++---- _news/announcement_1.md | 8 --- _news/announcement_2.md | 32 --------- _news/announcement_3.md | 8 --- _news/announcement_4.md | 8 +++ _news/announcement_5.md | 8 +++ _pages/about.md | 34 +++++++--- _pages/profiles.md | 4 +- _projects/10_project.md | 80 +++++++++++++++++++++++ _projects/11_project.md | 90 ++++++++++++++++++++++++++ _projects/12_project.md | 89 +++++++++++++++++++++++++ _projects/13_project.md | 100 ++++++++++++++++++++++++++++ _projects/14_project.md | 100 ++++++++++++++++++++++++++++ _projects/15_project.md | 109 +++++++++++++++++++++++++++++++ _projects/16_project.md | 118 +++++++++++++++++++++++++++++++++ _projects/1_project.md | 73 ++++++++++----------- _projects/2_project.md | 59 +++++++++++------ _projects/3_project.md | 60 +++++++++++------ _projects/4_project.md | 58 ++++++++++++----- _projects/5_project.md | 135 +++++++++++++++++--------------------- _projects/6_project.md | 140 +++++++++++++++++++--------------------- _projects/7_project.md | 62 ++++++++++++++++++ _projects/8_project.md | 71 ++++++++++++++++++++ _projects/9_project.md | 79 +++++++++++++++++++++++ docker-compose.yml | 1 - 27 files changed, 1292 insertions(+), 318 deletions(-) delete mode 100644 _news/announcement_1.md delete mode 100644 _news/announcement_2.md delete mode 100644 _news/announcement_3.md create mode 100644 _news/announcement_4.md create mode 100644 _news/announcement_5.md create mode 100644 _projects/10_project.md create mode 100644 _projects/11_project.md create mode 100644 _projects/12_project.md create mode 100644 _projects/13_project.md create mode 100644 _projects/14_project.md create mode 100644 _projects/15_project.md create mode 100644 _projects/16_project.md create mode 100644 _projects/7_project.md create mode 100644 _projects/8_project.md create mode 100644 _projects/9_project.md diff --git a/Gemfile b/Gemfile index 522f335..624c656 100644 --- a/Gemfile +++ b/Gemfile @@ -17,7 +17,6 @@ group :jekyll_plugins do gem 'jekyll-toc' gem 'jekyll-twitter-plugin' gem 'jemoji' - gem 'mini_racer' gem 'unicode_utils' gem 'webrick' end diff --git a/_config.yml b/_config.yml index 39d7dd7..977bbf5 100644 --- a/_config.yml +++ b/_config.yml @@ -8,11 +8,11 @@ middle_name: last_name: Kyamanywa email: untilhamza@gmail.com description: > # the ">" symbol means to ignore newlines until "footer_text:" - My personal site of Kyamanywa Hamza - full stack software engineer at Bebridge Inc, Seoul, South Korea. + Technical Founder of HoverNotes | Full-Stack + AI Engineer | Building privacy-first educational technology in Seoul, South Korea. footer_text: > keywords: > - Hamza, Hamza Kyamanywa, software engineer, Uganda , South Korea, Global Korea Scholarship, Javascript, Django, Langchain, NextJs, ReactJs, AWS, GCP # add your own keywords or leave empty + Hamza Kyamanywa, Technical Founder, HoverNotes, AI Engineer, Full-Stack Developer, Privacy-First EdTech, Educational Technology, Uganda, South Korea, Global Korea Scholarship, TypeScript, Python, React, Next.js, FastAPI, LangChain, LangGraph, Chrome Extension, Browser Extension Development, Explainable AI, Responsible AI, Machine Learning, Video Learning, Note-Taking, Obsidian, Local-First Architecture, Multi-modal AI, WebRTC, Real-time Processing, STT, Speech-to-Text, GPT-4, Claude, Gemini, Vector Databases, AWS, GCP, Docker, PostgreSQL, MongoDB, Redis, Springer Publication, Open Source, GitHub, Stack Overflow lang: en # the language of your site (for example: en, fr, cn, ru, etc.) icon: /favicon.ico # the emoji used as the favicon (alternatively, provide image name in /assets/img/) @@ -172,7 +172,7 @@ collections: permalink: /projects/:path/ announcements: - enabled: true + enabled: false scrollable: true # adds a vertical scroll bar if there are more than 3 news items limit: 5 # leave blank to include all the news in the `_news` folder diff --git a/_data/cv.yml b/_data/cv.yml index c1d3a66..2873c6a 100644 --- a/_data/cv.yml +++ b/_data/cv.yml @@ -6,7 +6,7 @@ - name: Date of Birth value: 30th October 1998 - name: Languages - value: English, Korea, Luganda + value: English, Korean, Luganda - name: Nationality value: Uganda @@ -19,7 +19,7 @@ description: - title: GPA contents: - - 4.02/4.5 - Magna Cum Laude + - 4.02/4.5 - Cum Laude - title: Relevant Coursework contents: - Data Structures and Algorithms @@ -57,27 +57,76 @@ - title: "Professional Experience" type: time_table contents: - - title: Software Engineer - institution: Bebridge Inc, Seoul South Korea - year: 2022 - present + - title: Technical Founder + institution: HoverNotes LLC + year: 2025 - present description: - - "Full stack development" - - "Backend API development" - - "Frontend app development" - - title: "Skills" - type: "list" + - "Building privacy-first AI video learning platform" + - "Universal video platform integration" + - "Chrome extension development with Manifest V3" + - "Local-first architecture with Obsidian integration" + - "45+ language support with cultural adaptations" + - title: "Technologies" contents: - - "Languages: Python, JavaScript" - - "Frameworks: Django, React, NextJS, Langchain" - - "Databases: MySQL, MongoDB, Firebase" - - "Other: Agile/SCRUM, Git, AWS, GCP" + - "Frontend: Next.js 15, React 19, TypeScript, Lexical Editor" + - "Backend: Python FastAPI, LangGraph, Docker, PostgreSQL" + - "AI/ML: Multi-modal processing, real-time transcription" + - "Browser: Chrome Extension APIs, File System Access API" + - title: Technical Lead (Freelance) + institution: GoGymi, Seoul South Korea + year: 2024 - 2025 + description: + - "Architecting Textutor AI essay correction platform" + - "GPT-4 integration with function calling" + - "Multi-language processing (5 languages)" + - "Real-time collaborative editing with Yjs" + - "Processing 1000+ daily submissions" + - title: "Technologies" + contents: + - "AI/ML: OpenAI GPT-4, Google Gemini, LanguageTool" + - "Frontend: Next.js 15, React 19, TypeScript" + - "Backend: FastAPI, PostgreSQL, Yjs" + - title: Principal AI Engineer / Technical Lead + institution: BEBRIDGE Inc (Slid), Seoul South Korea + year: 2022 - 2025 + description: + - "Pioneered Auto Notes System (60% cost optimization)" + - "Led Smart Live Text development (25% premium subscription increase)" + - "Built AI Sliddy chatbot and Universal Export system" + - "Developed multi-agent transcript correction systems" + - "YouTube URL extractor and remote capture innovations" + - title: "Key Technologies" + contents: + - "AI/ML: GPT-4, Claude, LangChain, LangGraph, Pinecone" + - "Frontend: React, Next.js, TypeScript, Redux" + - "Backend: Python, Node.js, FastAPI, WebSocket" + - "Infrastructure: AWS, Docker, Redis, FFmpeg" + +- title: Publications + type: time_table + contents: + - title: Responsible Music Genre Classification Using Interpretable Model-Agnostic Visual Explainers + year: 2025 + description: | + Co-authored paper published in SN Computer Science (Springer Nature). Advanced explainable AI in music genre classification using SHAP, LIME, and ELi5 techniques. + Developed responsible AI approach with CNN and Vision Transformer models achieving 80% accuracy with comprehensive model interpretability. + DOI: https://doi.org/10.1007/s42979-024-03584-9 - title: Open Source Contributions type: time_table contents: + - title: GitHub Leadership + year: 2019-2025 + description: 186+ public repositories, 535+ stars, Pull Shark (x4), Pair Extraordinaire (x2) achievements + - title: Stack Overflow Pioneer + year: 2022-2025 + description: First documented WebRTC β†’ Cloud Speech integration, solutions helping 5,000+ developers globally, Moderator election invitation - title: Langchain year: 2023 description: A framework for developing applications powered by language models (69.9 K stars). + - title: Quivr + year: 2024 + description: RAG AI integration framework for building AI assistants. - title: pytube year: 2023 description: A dependency-free Python library (and command-line utility) for downloading YouTube videos. (9.6 K stars). diff --git a/_news/announcement_1.md b/_news/announcement_1.md deleted file mode 100644 index e5349ce..0000000 --- a/_news/announcement_1.md +++ /dev/null @@ -1,8 +0,0 @@ ---- -layout: post -date: 2015-10-22 15:59:00-0400 -inline: true -related_posts: false ---- - -A simple inline announcement. diff --git a/_news/announcement_2.md b/_news/announcement_2.md deleted file mode 100644 index 9de9588..0000000 --- a/_news/announcement_2.md +++ /dev/null @@ -1,32 +0,0 @@ ---- -layout: post -title: A long announcement with details -date: 2015-11-07 16:11:00-0400 -inline: false -related_posts: false ---- - -Announcements and news can be much longer than just quick inline posts. In fact, they can have all the features available for the standard blog posts. See below. - -*** - -Jean shorts raw denim Vice normcore, art party High Life PBR skateboard stumptown vinyl kitsch. Four loko meh 8-bit, tousled banh mi tilde forage Schlitz dreamcatcher twee 3 wolf moon. Chambray asymmetrical paleo salvia, sartorial umami four loko master cleanse drinking vinegar brunch. Pinterest DIY authentic Schlitz, hoodie Intelligentsia butcher trust fund brunch shabby chic Kickstarter forage flexitarian. Direct trade cold-pressed meggings stumptown plaid, pop-up taxidermy. Hoodie XOXO fingerstache scenester Echo Park. Plaid ugh Wes Anderson, freegan pug selvage fanny pack leggings pickled food truck DIY irony Banksy. - -#### Hipster list - - -Hoodie Thundercats retro, tote bag 8-bit Godard craft beer gastropub. Truffaut Tumblr taxidermy, raw denim Kickstarter sartorial dreamcatcher. Quinoa chambray slow-carb salvia readymade, bicycle rights 90's yr typewriter selfies letterpress cardigan vegan. - -*** - -Pug heirloom High Life vinyl swag, single-origin coffee four dollar toast taxidermy reprehenderit fap distillery master cleanse locavore. Est anim sapiente leggings Brooklyn ea. Thundercats locavore excepteur veniam eiusmod. Raw denim Truffaut Schlitz, migas sapiente Portland VHS twee Bushwick Marfa typewriter retro id keytar. - -> We do not grow absolutely, chronologically. We grow sometimes in one dimension, and not in another, unevenly. We grow partially. We are relative. We are mature in one realm, childish in another. -> β€”Anais Nin - -Fap aliqua qui, scenester pug Echo Park polaroid irony shabby chic ex cardigan church-key Odd Future accusamus. Blog stumptown sartorial squid, gastropub duis aesthetic Truffaut vero. Pinterest tilde twee, odio mumblecore jean shorts lumbersexual. diff --git a/_news/announcement_3.md b/_news/announcement_3.md deleted file mode 100644 index 4d54088..0000000 --- a/_news/announcement_3.md +++ /dev/null @@ -1,8 +0,0 @@ ---- -layout: post -date: 2016-01-15 07:59:00-0400 -inline: true -related_posts: false ---- - -A simple inline announcement with Markdown emoji! :sparkles: :smile: diff --git a/_news/announcement_4.md b/_news/announcement_4.md new file mode 100644 index 0000000..4ab594f --- /dev/null +++ b/_news/announcement_4.md @@ -0,0 +1,8 @@ +--- +layout: post +date: 2025-01-01 12:00:00+0900 +inline: true +related_posts: false +--- + +Founded [HoverNotes LLC](https://hovernotes.app), a privacy-first AI video learning platform that eliminates context-switching in video-based learning. πŸš€ \ No newline at end of file diff --git a/_news/announcement_5.md b/_news/announcement_5.md new file mode 100644 index 0000000..087cd60 --- /dev/null +++ b/_news/announcement_5.md @@ -0,0 +1,8 @@ +--- +layout: post +date: 2025-01-15 10:00:00+0900 +inline: true +related_posts: false +--- + +Published research paper "[Responsible Music Genre Classification Using Interpretable Model-Agnostic Visual Explainers](https://doi.org/10.1007/s42979-024-03584-9)" in SN Computer Science (Springer Nature). πŸ“š \ No newline at end of file diff --git a/_pages/about.md b/_pages/about.md index a0b0a52..d6ccc9d 100644 --- a/_pages/about.md +++ b/_pages/about.md @@ -6,11 +6,11 @@ subtitle: #Affiliations. Address. Contacts. Moto. Etc. profile: align: right - image: prof_pic.JPG + image: prof_pic_2.JPG image_circular: false # crops the image to make it circular more_info: > -

Software Engineer

-

Bebridge Inc

+

Technical Founder @ HoverNotes

+

AI Lead @ GoGymi

Seoul, South Korea

news: false # includes a list of news items @@ -40,11 +40,20 @@ In my free time, you will find me either coding, playing guitar, playing FIFA, o #### Welcome to My Personal Website -I am a Ugandan a full-stack software engineer and AI enthusiast currently living and working in Seoul, South Korea. +I am a Technical Founder and AI Engineer specializing in full-stack + AI development with 4+ years of hands-on experience building revolutionary educational technology. Currently founding HoverNotes while freelancing as Technical Lead at GoGymi. #### Education & Professional Career -I am a proud recipient of the Ugandan Government Scholarship (2017) and the [Global Korea Scholarship](http://www.niied.go.kr/user/nd35203.do) (2018 - 2023). I graduated magna cum laude in Electrical Engineering from [Korea University](https://www.korea.edu/mbshome/mbs/en/index.do), Seoul South Korea (class of 2023). My professional journey has led me to [Bebridge Inc](https://home.slid.cc), where I blend technology and creativity. I am also a volunteer AI researcher with Neural Securify AI Group, where I dive deep into Generative AI and Responsible AI. +I am a proud recipient of the Ugandan Government Scholarship (2017) and the [Global Korea Scholarship](http://www.niied.go.kr/user/nd35203.do) (2018 - 2023). I graduated cum laude in Electrical Engineering from [Korea University](https://www.korea.edu/mbshome/mbs/en/index.do), Seoul South Korea (class of 2023) with a 4.02/4.5 GPA. + +**Current Roles:** +- **Technical Founder @ HoverNotes**: Building privacy-first AI video learning platform with universal video integration, local Obsidian storage, and 45+ language support +- **AI Lead @ GoGymi**: Architecting Textutor AI platform for essay correction and grading, processing 1000+ daily submissions + +**Previous Experience:** +- **Principal AI Engineer @ Slid (2022-2025)**: Pioneered revolutionary features including Auto Notes System (60% cost optimization), Smart Live Text (25% premium subscription increase), AI Sliddy chatbot, and multi-agent transcript correction systems + +**Research:** Co-authored paper in SN Computer Science (Springer Nature) on responsible AI and explainable machine learning. @@ -115,11 +124,16 @@ Interested in LLMs, software innovation, or just a friendly chat about the lates ### Skills -- **Languages:** Python, JavaScript, C++ -- **Frameworks:** Django, React, NextJS, Flask, Bootstrap, NestJs, Tailwind -- **Machine Learning:** Langchain, Pytorch, Keras -- **Databases:**Β MySQL, MongoDB, Firebase, Postgress, Supabase -- **Other:** Agile/SCRUM, Git, AWS, GCP +- **Languages:** TypeScript, Python, JavaScript, C++ +- **Frontend:** React 19, Next.js 15, React Native, Chrome Extensions (Manifest V3), Tailwind CSS, Bootstrap +- **Backend:** FastAPI, Node.js, NestJS, Express, Django, Flask +- **AI/ML:** LangChain, LangGraph, GPT-4, Claude, Gemini, PyTorch, Keras, SHAP, LIME, Vector Databases +- **Real-time:** WebRTC, WebSocket, Speech-to-Text (STT), Real-time Transcription +- **Databases:**Β PostgreSQL, MongoDB, Redis, MySQL, Firebase, Supabase, Vector DBs +- **Infrastructure:** Docker, Kubernetes, AWS, GCP, Multi-region Deployment, CDN +- **Browser APIs:** File System Access API, Clipboard API, Screen Capture API, Media APIs +- **Architecture:** Microservices, Multi-agent Systems, Event-driven, Serverless +- **Other:** Git, Agile/SCRUM, Anti-detection Systems, Performance Optimization
diff --git a/_pages/profiles.md b/_pages/profiles.md index 4982d65..341b82c 100644 --- a/_pages/profiles.md +++ b/_pages/profiles.md @@ -10,7 +10,7 @@ profiles: # if you want to include more than one profile, just replicate the following block # and create one content file for each profile inside _pages/ - align: right - image: prof_pic.jpg + image: prof_pic_2.jpg content: about_einstein.md image_circular: false # crops the image to make it circular more_info: > @@ -18,7 +18,7 @@ profiles:

123 your address street

Your City, State 12345

- align: left - image: prof_pic.jpg + image: prof_pic_2.jpg content: about_einstein.md image_circular: false # crops the image to make it circular more_info: > diff --git a/_projects/10_project.md b/_projects/10_project.md new file mode 100644 index 0000000..f0fb0c3 --- /dev/null +++ b/_projects/10_project.md @@ -0,0 +1,80 @@ +--- +layout: page +title: Video Integration & Media Management +description: Universal platform support with seamless video integration across all platforms +img: +importance: 10 +category: work +--- + +
+

Description:

+

Built comprehensive video integration supporting YouTube, Vimeo, and custom video platforms with unified video interface abstracting platform differences and security boundary handling. Implemented sophisticated media permission orchestration across Chrome, Firefox, and Safari with sub-100ms content rendering.

+ +

Project Duration:

+

2022-2024 (Full Stack Engineer at Slid)

+ +

Key Technical Achievements:

+ + +

Platform Integrations:

+ + +

Technical Implementation:

+ + +

Cross-Browser Compatibility:

+ + +

Media Permission System:

+ + +

Business Impact:

+ + +

Technical Innovations:

+ + +

Skills Demonstrated:

+

Video APIs, Cross-Browser Development, Media Permissions, Performance Optimization, API Integration, Security, React Development

+
\ No newline at end of file diff --git a/_projects/11_project.md b/_projects/11_project.md new file mode 100644 index 0000000..1d16ec9 --- /dev/null +++ b/_projects/11_project.md @@ -0,0 +1,90 @@ +--- +layout: page +title: Payment & Pricing System +description: Multi-gateway payment system with subscription management and privilege control +img: +importance: 11 +category: work +--- + +
+

Description:

+

Implemented comprehensive multi-gateway payment system with Stripe (international) and Payple (Korean) integration, featuring subscription management, innovative Day Pass system, and granular privilege management. Achieved 35% improvement in checkout completion and 98%+ payment success rate.

+ +

Project Duration:

+

2022-2023 (Full Stack Engineer at Slid)

+ +

Key Technical Achievements:

+ + +

Payment Gateway Integration:

+ + +

Subscription Management:

+ + +

Technical Implementation:

+ + +

Access Control System:

+ + +

Day Pass Innovation:

+ + +

Business Impact:

+ + +

Technical Innovations:

+ + +

Skills Demonstrated:

+

Payment Processing, Subscription Management, API Integration, Security, Database Design, Business Logic, Multi-Platform Development

+
\ No newline at end of file diff --git a/_projects/12_project.md b/_projects/12_project.md new file mode 100644 index 0000000..ac32807 --- /dev/null +++ b/_projects/12_project.md @@ -0,0 +1,89 @@ +--- +layout: page +title: Editor & Block System Innovation +description: React-EditorJS integration breakthrough replacing vanilla JavaScript implementations +img: +importance: 12 +category: work +--- + +
+

Description:

+

Pioneered integration of React components within EditorJS block system, replacing vanilla JavaScript implementations with modern React architecture. Built sophisticated AST processing supporting complex formatting, lazy loading, and virtualized rendering that became the foundation for Auto Notes and Smart Live Text features.

+ +

Project Duration:

+

2022-2024 (Full Stack Engineer at Slid)

+ +

Key Technical Achievements:

+ + +

Technical Breakthrough Details:

+ + +

Advanced Features:

+ + +

Technical Implementation:

+ + +

AST & Content Processing:

+ + +

Integration Ecosystem:

+ + +

Business Impact:

+ + +

Technical Innovations:

+ + +

Skills Demonstrated:

+

React Development, EditorJS, AST Processing, Performance Optimization, System Architecture, TypeScript, Content Management, UI/UX Design

+
\ No newline at end of file diff --git a/_projects/13_project.md b/_projects/13_project.md new file mode 100644 index 0000000..b6062b9 --- /dev/null +++ b/_projects/13_project.md @@ -0,0 +1,100 @@ +--- +layout: page +title: Authentication Architecture 2.0 +description: Advanced auth system with AWS Cognito extension for multi-platform support +img: +importance: 13 +category: work +--- + +
+

Description:

+

Led architecture evolution from basic token authentication to sophisticated Auth 2.0 with cookie-based security. Extended AWS Cognito SDK with CustomCognitoStorage for multi-platform support across web, extension, desktop, and mobile contexts, implementing granular privilege management with zero unauthorized access.

+ +

Project Duration:

+

2022-2023 (Full Stack Engineer at Slid)

+ +

Key Technical Achievements:

+ + +

Multi-Platform Architecture:

+ + +

CustomCognitoStorage Innovation:

+ + +

Technical Implementation:

+ + +

Access Control System:

+ + +

Security Measures:

+ + +

Migration Strategy:

+ + +

Business Impact:

+ + +

Technical Innovations:

+ + +

Skills Demonstrated:

+

AWS Cognito, OAuth 2.0, JWT, Security Architecture, Multi-Platform Development, System Migration, Access Control, Compliance

+
\ No newline at end of file diff --git a/_projects/14_project.md b/_projects/14_project.md new file mode 100644 index 0000000..93185ad --- /dev/null +++ b/_projects/14_project.md @@ -0,0 +1,100 @@ +--- +layout: page +title: Extension Communication System +description: Sophisticated cross-platform messaging between browser extension, web app, and desktop +img: +importance: 14 +category: work +--- + +
+

Description:

+

Built sophisticated communication system enabling seamless interaction between browser extension, web application, and desktop application. Designed 43+ message types for video capture, permissions, and state synchronization with pixel-perfect capture area sync and real-time control.

+ +

Project Duration:

+

2022-2024 (Full Stack Engineer at Slid)

+ +

Key Technical Achievements:

+ + +

Communication Architecture:

+ + +

Message Types & Protocols:

+ + +

Technical Implementation:

+ + +

Capture Area Synchronization:

+ + +

State Management System:

+ + +

Security & Validation:

+ + +

Business Impact:

+ + +

Technical Innovations:

+ + +

Skills Demonstrated:

+

Chrome Extension APIs, WebSocket, Cross-Platform Communication, Real-time Systems, Canvas Processing, State Management, Security

+
\ No newline at end of file diff --git a/_projects/15_project.md b/_projects/15_project.md new file mode 100644 index 0000000..11f1f0a --- /dev/null +++ b/_projects/15_project.md @@ -0,0 +1,109 @@ +--- +layout: page +title: Mobile App Innovation +description: Overcome platform limitations through sophisticated backend video processing +img: +importance: 15 +category: work +--- + +
+

Description:

+

Solved fundamental mobile platform challenge by building sophisticated backend architecture enabling mobile users to capture precise video screenshots. Developed React Native frontend with Python FastAPI backend for video stream processing, featuring anti-detection systems and sub-second processing with global CDN distribution.

+ +

Project Duration:

+

2022-2024 (Full Stack Engineer at Slid)

+ +

Key Technical Achievements:

+ + +

Mobile Platform Challenges Solved:

+ + +

Technical Architecture:

+ + +

Backend Processing Pipeline:

+ + +

Mobile App Features:

+ + +

Anti-Detection & Reliability:

+ + +

Performance Optimizations:

+ + +

Global Infrastructure:

+ + +

Business Impact:

+ + +

Technical Innovations:

+ + +

Skills Demonstrated:

+

React Native, Mobile Development, Backend Architecture, Video Processing, Anti-Detection Systems, Global Infrastructure, Performance Optimization

+
\ No newline at end of file diff --git a/_projects/16_project.md b/_projects/16_project.md new file mode 100644 index 0000000..bfe3229 --- /dev/null +++ b/_projects/16_project.md @@ -0,0 +1,118 @@ +--- +layout: page +title: Real-Time Video Processing Infrastructure +description: Production-scale video processing with sub-second latency and anti-detection systems +img: +importance: 16 +category: work +--- + +
+

Description:

+

Built production-scale video processing infrastructure with sub-second latency for educational content. Implemented advanced anti-detection systems with IP rotation and user agent switching, supporting concurrent processing for thousands of educational users with multi-region deployment and intelligent caching.

+ +

Project Duration:

+

2022-2024 (Full Stack Engineer at Slid)

+ +

Key Technical Achievements:

+ + +

Infrastructure Architecture:

+ + +

Real-Time Processing Pipeline:

+ + +

Technical Stack:

+ + +

Anti-Detection Systems:

+ + +

Performance Optimizations:

+ + +

Multi-Region Deployment:

+ + +

Monitoring & Analytics:

+ + +

Security & Compliance:

+ + +

Business Impact:

+ + +

Technical Innovations:

+ + +

Skills Demonstrated:

+

Infrastructure Design, Video Processing, Docker, Kubernetes, Performance Optimization, Anti-Detection Systems, Global Deployment, System Monitoring

+
\ No newline at end of file diff --git a/_projects/1_project.md b/_projects/1_project.md index 6a66e92..3bf731d 100644 --- a/_projects/1_project.md +++ b/_projects/1_project.md @@ -1,50 +1,51 @@ --- layout: page -title: Smart Live Text -description: Led the development of an advanced real-time transcription feature +title: HoverNotes +description: Privacy-first AI video learning platform eliminating context-switching img: # assets/img/12.jpg importance: 1 category: work -# related_publications: einstein1956investigations, einstein1950meaning -# NOTE: projects can have related publications, so think about it when you work on projects at Slid --- - -

Description:

-

Led the development of an advanced real-time transcription feature called "Smart Live Text" for use with Slid's note-taking product. The feature enables users to easily take notes while watching videos in our Chrome extension, with a 85% accuracy rate.

+

Founded and leading technical development of HoverNotes, a groundbreaking privacy-first AI video note-taking platform that eliminates context-switching in video-based learning. The platform features a sophisticated three-tier architecture (browser extension + web app + AI backend) enabling note-taking on any video platform while maintaining complete user privacy through local Obsidian vault storage.

+

Project Duration:

-

November 2022– December 2022

-

Programming Commitment:

-

Full-time (4 hours coding, 2 hours maintenance, 2 hours testing per day, Monday to Friday)

-

Key Responsibilities:

+

January 2025 – Present (Founder & Technical Lead)

+ +

Key Technical Achievements:

+ + +

Technical Stack:

+ + +

Business Impact:

-

Achievements:

+ +

Innovation Highlights:

-
+ \ No newline at end of file diff --git a/_projects/2_project.md b/_projects/2_project.md index 7638b7a..12a57ae 100644 --- a/_projects/2_project.md +++ b/_projects/2_project.md @@ -1,31 +1,54 @@ --- layout: page -title: Slid Authentication 2.0 -description: Led the overhaul of the platform's authentication system -img: # assets/img/3.jpg +title: Auto Notes System +description: Pioneered AI-powered auto note generation from video content at Slid +img: # assets/img/12.jpg importance: 2 category: work -giscus_comments: false ---

Description:

-

Led the overhaul of the platform's authentication system to address constant issues and increase user trust.

+

Pioneered the development of AI-powered auto note generation from video content at Slid, starting as a side project before the AI boom. Built an end-to-end system that transforms video content into structured, actionable notes using advanced AI processing, achieving significant cost optimization and establishing new revenue streams.

+

Project Duration:

-

October 2022 – November 2022

-

Programming Commitment:

-

Full-time (4 hours coding, 2 hours maintenance, 2 hours testing per day, Monday to Friday)

-

Key Responsibilities:

+

2024 (Principal AI Engineer at Slid)

+ +

Key Technical Achievements:

-

Achievements:

+ +

Technical Implementation:

-
+ +

Business Impact:

+ + +

Technical Innovations:

+ + +

Skills Demonstrated:

+

AI/ML Engineering, System Architecture, Cost Optimization, Real-time Processing, API Integration, Product Innovation, Full-stack Development

+ \ No newline at end of file diff --git a/_projects/3_project.md b/_projects/3_project.md index eccc340..63e6fe8 100644 --- a/_projects/3_project.md +++ b/_projects/3_project.md @@ -1,32 +1,54 @@ --- layout: page -title: Universal Note Copy and Paste -description: Led the development of a universal exportation feature for Slid's note-taking product. +title: Textutor AI Platform +description: Led technical architecture for AI-powered essay correction and grading system at GoGymi img: # assets/img/7.jpg -# redirect: https://unsplash.com importance: 3 category: work --- -
+

Description:

-

Led the development of a universal exportation feature for Slid's note-taking product, allowing users to easily copy and paste their notes to any platform, including Medium, Notion, Microsoft Word, Visual Studio Code, and Apple Notes. The feature processed over 35,000 data points per week and achieved a 95% accuracy rate in conversion to various formats.

+

Led technical architecture for Textutor AI Platform at GoGymi - a comprehensive AI-powered essay correction and grading system revolutionizing language education. Built sophisticated GPT-4 integration with function calling for structured educational assessment, processing 1000+ student submissions daily with 99.9% uptime.

+

Project Duration:

-

October 2022 – November 2022

-

Programming Commitment:

-

Full-time (4 hours coding, 2 hours maintenance, 2 hours testing per day, Monday to Friday)

-

Key Responsibilities:

+

2024 - 2025 (Technical Lead, Freelance)

+ +

Key Technical Achievements:

    -
  • Implemented a new solution using the browser clipboard API to paste Slid notes in different formats to various platforms, processing over 5,000 data points per day with a 95% accuracy rate.
    • -
  • Researched and learned how to work with the clipboard API and hold multiple data formats on the clipboard at the same time.
    • -
  • Developed a custom clipboard class and interface due to compatibility issues with our version of TypeScript.
    • -
  • Created a performant process for converting editor data into HTML format on the fly.
    • -
  • Managed the project using tools such as Visual Studio Code, Git, GitHub, Sentry, Notion, and Slack.
    • +
  • AI Integration Excellence: Built sophisticated GPT-4 integration with function calling for structured educational assessment
    • +
  • Multi-Language Processing: Implemented LanguageTool integration supporting English, German, Spanish, French, and Italian with cultural adaptations
    • +
  • Real-Time Collaboration: Architected Yjs-powered collaborative editing with live feedback, comments, and presence indicators
    • +
  • Database Architecture: Designed complex 16-table PostgreSQL schema with audit trails, versioning, and educational analytics
    • +
  • Performance Optimization: Achieved sub-10 second comprehensive essay analysis with 99.9% uptime
    • +
  • Hybrid Grading System: Implemented AI-human hybrid grading workflow for quality assurance
-

Achievements:

+ +

Technical Implementation:

    -
  • Successfully developed and deployed the Universal Note Copy and Paste feature, providing Slid users with increased flexibility and convenience in sharing their notes on other platforms.
    • -
  • Improved user experience and satisfaction by providing a highly requested feature that had previously been a limitation for Slid users, resulting in a 20% increase in positive feedback and a 25% decrease in support requests related to copy and paste functionality.
    • -
  • Demonstrated strong technical skills and problem-solving abilities in overcoming complex challenges and continuously improving the feature after deployment.
    • +
  • AI/ML Stack: OpenAI GPT-4 with function calling, Google Gemini, Custom prompt engineering
    • +
  • Frontend: Next.js 15, React 19, TypeScript, Tailwind CSS, Yjs for collaboration
    • +
  • Backend: Python FastAPI, PostgreSQL with Drizzle ORM, Redis for caching
    • +
  • Language Processing: LanguageTool API integration, Custom grammar rules, Multi-language support
    • +
  • Real-time Features: WebSocket connections, Presence tracking, Live cursor positions
-
+ +

Business Impact:

+ + +

Technical Innovations:

+ + +

Skills Demonstrated:

+

Technical Leadership, AI/ML Integration, Database Design, Real-time Systems, Educational Technology, Multi-language Processing, System Architecture, Performance Optimization

+
\ No newline at end of file diff --git a/_projects/4_project.md b/_projects/4_project.md index 0ca1b11..a73e0ae 100644 --- a/_projects/4_project.md +++ b/_projects/4_project.md @@ -1,7 +1,7 @@ --- layout: page -title: Advanced User Behavior Monitoring System -description: Implemented a sophisticated user behavior monitoring system +title: AI Sliddy Chat +description: Built industry-first AI chat for notes years before Notion AI and competitors img: importance: 4 category: work @@ -9,21 +9,49 @@ category: work

Description:

-

Developed and implemented a sophisticated user behavior monitoring system using React hooks to track and analyze user activity on our site.

+

Led the development of AI Sliddy, an innovative chatbot that integrated GPT models into Slid's note-taking product years before Notion AI and other competitors entered the market. Using LangChain, Pinecone vector DB, and OpenAI's GPT API, enabled users to chat and ask questions about their notes content, resulting in significant engagement increases.

+

Project Duration:

-

September 2022 - September 2022

-

Programming Commitment:

-

Full-time (4 hours coding, 2 hours maintenance, 2 hours testing per day, Monday to Friday)

-

Responsibilities:

+

2023 (Senior Software Engineer at Slid)

+ +

Key Technical Achievements:

-

Achievements:

+ +

Technical Implementation:

-
+ +

Business Impact:

+ + +

Technical Innovations:

+ + +

Open Source Contributions:

+

Contributed to Vercel AI SDK development based on learnings from this project, helping shape the future of AI development tools for the broader community.

+ +

Skills Demonstrated:

+

AI/ML Engineering, Vector Databases, Semantic Search, Real-time Systems, API Design, Product Innovation, Open Source Contribution

+ \ No newline at end of file diff --git a/_projects/5_project.md b/_projects/5_project.md index efd9b6c..0e34c35 100644 --- a/_projects/5_project.md +++ b/_projects/5_project.md @@ -1,80 +1,63 @@ --- layout: page -title: project 5 -description: a project with a background image -img: assets/img/1.jpg -importance: 3 -category: fun +title: Universal Export System +description: Built comprehensive multi-format export system reducing support tickets by 75% +img: #assets/img/1.jpg +importance: 5 +category: work --- -Every project has a beautiful feature showcase page. -It's easy to include images in a flexible 3-column grid format. -Make your photos 1/3, 2/3, or full width. - -To give your project a background in the portfolio page, just add the img tag to the front matter like so: - - --- - layout: page - title: project - description: a project with a background image - img: /assets/img/12.jpg - --- - -
-
- {% include figure.html path="assets/img/1.jpg" title="example image" class="img-fluid rounded z-depth-1" %} -
-
- {% include figure.html path="assets/img/3.jpg" title="example image" class="img-fluid rounded z-depth-1" %} -
-
- {% include figure.html path="assets/img/5.jpg" title="example image" class="img-fluid rounded z-depth-1" %} -
-
-
- Caption photos easily. On the left, a road goes through a tunnel. Middle, leaves artistically fall in a hipster photoshoot. Right, in another hipster photoshoot, a lumberjack grasps a handful of pine needles. -
-
-
- {% include figure.html path="assets/img/5.jpg" title="example image" class="img-fluid rounded z-depth-1" %} -
-
-
- This image can also have a caption. It's like magic. -
- -You can also put regular text between your rows of images. -Say you wanted to write a little bit about your project before you posted the rest of the images. -You describe how you toiled, sweated, *bled* for your project, and then... you reveal its glory in the next row of images. - - -
-
- {% include figure.html path="assets/img/6.jpg" title="example image" class="img-fluid rounded z-depth-1" %} -
-
- {% include figure.html path="assets/img/11.jpg" title="example image" class="img-fluid rounded z-depth-1" %} -
-
-
- You can also have artistically styled 2/3 + 1/3 images, like these. -
- - -The code is simple. -Just wrap your images with `
` and place them inside `
` (read more about the Bootstrap Grid system). -To make images responsive, add `img-fluid` class to each; for rounded corners and shadows use `rounded` and `z-depth-1` classes. -Here's the code for the last row of images above: - -{% raw %} -```html -
-
- {% include figure.html path="assets/img/6.jpg" title="example image" class="img-fluid rounded z-depth-1" %} -
-
- {% include figure.html path="assets/img/11.jpg" title="example image" class="img-fluid rounded z-depth-1" %} -
-
-``` -{% endraw %} +
+

Description:

+

Built a comprehensive multi-format export system for Slid that allowed users to export their notes to Word, PDF, Markdown, images, and Notion clipboard integration. The system featured advanced content parsing handling nested structures, media assets, and cross-platform compatibility, achieving a 75% reduction in export-related support tickets.

+ +

Project Duration:

+

2023 (Senior Software Engineer at Slid)

+ +

Key Technical Achievements:

+
    +
  • Multi-Format Support: Implemented export to Word, PDF, Markdown, HTML, and image formats
    • +
  • Notion Integration: Custom clipboard API supporting simultaneous HTML and plain text for intelligent paste behavior
    • +
  • Advanced Parsing: Built sophisticated content parser handling nested structures and media assets
    • +
  • Cross-Platform Compatibility: Ensured seamless export across Windows, macOS, and Linux
    • +
  • Mobile Optimization: Adapted export functionality for mobile platforms with size constraints
    • +
  • Performance: Optimized large document exports handling 100+ page documents efficiently
    • +
+ +

Technical Implementation:

+
    +
  • Core Architecture: GlobalExport class with format-specific renderers
    • +
  • Document Generation: Custom libraries for DOCX, PDF generation with styling preservation
    • +
  • Clipboard Engineering: Browser Clipboard API with multi-format data handling
    • +
  • Frontend: React components with export progress tracking and preview
    • +
  • Backend: Node.js with Express, file streaming for large exports
    • +
+ +

Format-Specific Features:

+
    +
  • Word Export: Full formatting preservation, embedded images, table support
    • +
  • PDF Generation: Custom styling, page breaks, header/footer support
    • +
  • Markdown Conversion: Clean markdown with front matter and metadata
    • +
  • Notion Clipboard: Smart paste with block type detection and formatting
    • +
  • Image Export: High-resolution screenshots with customizable dimensions
    • +
+ +

Business Impact:

+
    +
  • Support Reduction: 75% decrease in export-related support tickets
    • +
  • User Satisfaction: Export feature became top-rated feature in user surveys
    • +
  • Platform Stickiness: Seamless content portability increased user retention
    • +
  • Market Differentiation: Best-in-class export capabilities vs competitors
    • +
+ +

Technical Innovations:

+
    +
  • Smart Format Detection: Automatic optimal format suggestion based on content type
    • +
  • Incremental Export: Streaming architecture for memory-efficient large exports
    • +
  • Format Preservation: Advanced AST manipulation maintaining formatting fidelity
    • +
  • Error Recovery: Graceful handling of corrupt content and media assets
    • +
+ +

Skills Demonstrated:

+

System Architecture, Document Processing, Cross-Platform Development, API Design, Performance Optimization, User Experience Design

+
\ No newline at end of file diff --git a/_projects/6_project.md b/_projects/6_project.md index 9a95d6e..b85f848 100644 --- a/_projects/6_project.md +++ b/_projects/6_project.md @@ -1,80 +1,70 @@ --- layout: page -title: project 6 -description: a project with no image +title: Smart Live Text (STT Pioneer) +description: First successful WebRTC β†’ Google Cloud Speech implementation helping 5,000+ developers img: -importance: 4 -category: fun +importance: 6 +category: work --- -Every project has a beautiful feature showcase page. -It's easy to include images in a flexible 3-column grid format. -Make your photos 1/3, 2/3, or full width. - -To give your project a background in the portfolio page, just add the img tag to the front matter like so: - - --- - layout: page - title: project - description: a project with a background image - img: /assets/img/12.jpg - --- - -
-
- {% include figure.html path="assets/img/1.jpg" title="example image" class="img-fluid rounded z-depth-1" %} -
-
- {% include figure.html path="assets/img/3.jpg" title="example image" class="img-fluid rounded z-depth-1" %} -
-
- {% include figure.html path="assets/img/5.jpg" title="example image" class="img-fluid rounded z-depth-1" %} -
-
-
- Caption photos easily. On the left, a road goes through a tunnel. Middle, leaves artistically fall in a hipster photoshoot. Right, in another hipster photoshoot, a lumberjack grasps a handful of pine needles. -
-
-
- {% include figure.html path="assets/img/5.jpg" title="example image" class="img-fluid rounded z-depth-1" %} -
-
-
- This image can also have a caption. It's like magic. -
- -You can also put regular text between your rows of images. -Say you wanted to write a little bit about your project before you posted the rest of the images. -You describe how you toiled, sweated, *bled* for your project, and then... you reveal its glory in the next row of images. - - -
-
- {% include figure.html path="assets/img/6.jpg" title="example image" class="img-fluid rounded z-depth-1" %} -
-
- {% include figure.html path="assets/img/11.jpg" title="example image" class="img-fluid rounded z-depth-1" %} -
-
-
- You can also have artistically styled 2/3 + 1/3 images, like these. -
- - -The code is simple. -Just wrap your images with `
` and place them inside `
` (read more about the Bootstrap Grid system). -To make images responsive, add `img-fluid` class to each; for rounded corners and shadows use `rounded` and `z-depth-1` classes. -Here's the code for the last row of images above: - -{% raw %} -```html -
-
- {% include figure.html path="assets/img/6.jpg" title="example image" class="img-fluid rounded z-depth-1" %} -
-
- {% include figure.html path="assets/img/11.jpg" title="example image" class="img-fluid rounded z-depth-1" %} -
-
-``` -{% endraw %} +
+

Description:

+

Pioneered the first successful WebRTC β†’ Socket.io β†’ Google Cloud Speech streaming implementation for real-time video transcription. This breakthrough solution became a Stack Overflow reference helping 5,000+ developers worldwide, established Slid as an industry leader in video note-taking, and resulted in a 25% increase in premium subscriptions.

+ +

Project Duration:

+

2022-2023 (Full Stack Engineer at Slid)

+ +

Key Technical Achievements:

+
    +
  • WebRTC Pioneer: First documented successful implementation of WebRTC to Cloud Speech streaming
    • +
  • Stack Overflow Impact: Solution became canonical reference for real-time STT implementation
    • +
  • Cost Optimization Journey: 90% cost reduction through provider evolution (Whisper β†’ Google β†’ Groq)
    • +
  • AudioWorklet Innovation: Custom AudioWorkletProcessor with real-time downsampling (44.1kHz β†’ 16kHz)
    • +
  • Cross-Browser Compatibility: Unified implementation across Chrome, Firefox, and Safari
    • +
  • Production Scale: Handling thousands of concurrent transcription sessions
    • +
+ +

Technical Implementation:

+
    +
  • Audio Pipeline: WebRTC β†’ AudioWorkletProcessor β†’ Socket.io β†’ Server β†’ Google Cloud Speech
    • +
  • Frontend: React with custom audio processing hooks, Redux for state management
    • +
  • Backend: Node.js with Socket.io, Google Cloud Speech API integration
    • +
  • Audio Processing: Real-time resampling, noise reduction, silence detection
    • +
  • Optimization: Adaptive bitrate, intelligent buffering, connection pooling
    • +
+ +

Community Contribution:

+
    +
  • Stack Overflow Solution: Published comprehensive solution
    • +
  • Developer Impact: 5,000+ developers helped, 100+ implementations based on solution
    • +
  • Documentation: Created detailed implementation guide with code examples
    • +
  • Community Support: Ongoing assistance to developers implementing similar solutions
    • +
+ +

Cost Optimization Evolution:

+
    +
  • Phase 1 - OpenAI Whisper: Initial implementation, high accuracy but expensive
    • +
  • Phase 2 - Google Cloud Speech: 70% cost reduction with comparable accuracy
    • +
  • Phase 3 - Groq Integration: 90% total cost reduction with improved latency
    • +
  • Smart Routing: Automatic provider selection based on language and quality requirements
    • +
+ +

Business Impact:

+
    +
  • Revenue Growth: 25% increase in premium subscriptions
    • +
  • Market Leadership: Established Slid as STT innovation leader
    • +
  • User Retention: 40% improvement in user engagement metrics
    • +
  • Cost Efficiency: 90% reduction in transcription costs while improving quality
    • +
+ +

Technical Innovations:

+
    +
  • Adaptive Streaming: Dynamic quality adjustment based on network conditions
    • +
  • Language Detection: Automatic language identification for optimal model selection
    • +
  • Error Recovery: Graceful reconnection with minimal transcript loss
    • +
  • Privacy Protection: End-to-end encryption for sensitive audio streams
    • +
+ +

Skills Demonstrated:

+

WebRTC, Real-time Audio Processing, Cloud Integration, Cost Optimization, Community Leadership, Technical Documentation, Production Scaling

+
\ No newline at end of file diff --git a/_projects/7_project.md b/_projects/7_project.md new file mode 100644 index 0000000..7afb387 --- /dev/null +++ b/_projects/7_project.md @@ -0,0 +1,62 @@ +--- +layout: page +title: YouTube URL Extractor +description: Reverse-engineered YouTube's video delivery for direct server URL access +img: +importance: 7 +category: work +--- + +
+

Description:

+

Discovered and implemented a method to extract direct YouTube video URLs, enabling server-side processing for mobile platforms. This breakthrough became the foundation for multiple features including mobile video processing and remote capture systems, supporting thousands of concurrent educational requests.

+ +

Project Duration:

+

2024 (Principal AI Engineer at Slid)

+ +

Key Technical Achievements:

+
    +
  • Reverse Engineering: Discovered method to extract direct YouTube video URLs for educational content access
    • +
  • Foundation Impact: Discovery became backbone for mobile app video processing and remote capture systems
    • +
  • Production Scale: Robust system supporting thousands of concurrent educational requests
    • +
  • Anti-Detection Systems: Implemented sophisticated mechanisms to ensure reliable access
    • +
  • Cross-Platform Bridge: Enabled desktop-quality features on mobile platforms
    • +
  • Performance Optimization: Sub-second URL extraction with intelligent caching
    • +
+ +

Technical Implementation:

+
    +
  • Core Technology: Python with yt-dlp integration and custom parsing algorithms
    • +
  • URL Processing: Format optimization, quality selection, adaptive bitrate handling
    • +
  • Anti-Detection: IP rotation, user agent management, request throttling
    • +
  • Caching Layer: Redis-based caching for frequently accessed content
    • +
  • API Design: RESTful endpoints with authentication and rate limiting
    • +
+ +

System Architecture:

+
    +
  • Discovery Service: Automated URL extraction with fallback mechanisms
    • +
  • Validation Pipeline: URL verification and quality assurance
    • +
  • Distribution Network: Load-balanced servers across multiple regions
    • +
  • Monitoring: Real-time health checks and automatic failover
    • +
+ +

Business Impact:

+
    +
  • Mobile Enablement: Unlocked video processing capabilities for mobile users
    • +
  • Feature Foundation: Enabled development of remote capture and processing systems
    • +
  • User Reach: Extended platform capabilities to mobile-first markets
    • +
  • Competitive Advantage: Unique capability differentiating from competitors
    • +
+ +

Technical Innovations:

+
    +
  • Adaptive Extraction: Dynamic algorithm adjustment based on platform changes
    • +
  • Quality Optimization: Intelligent format selection for bandwidth efficiency
    • +
  • Resilience Design: Multi-layer fallback ensuring 99.9% availability
    • +
  • Security Measures: Encrypted communication and access control
    • +
+ +

Skills Demonstrated:

+

Reverse Engineering, System Architecture, Python Development, API Design, Anti-Detection Systems, Performance Optimization, Production Scaling

+
\ No newline at end of file diff --git a/_projects/8_project.md b/_projects/8_project.md new file mode 100644 index 0000000..94945e6 --- /dev/null +++ b/_projects/8_project.md @@ -0,0 +1,71 @@ +--- +layout: page +title: Remote Capture System +description: Server-side video screenshot system enabling mobile video processing +img: +importance: 8 +category: work +--- + +
+

Description:

+

Built sophisticated remote screenshot capture system leveraging the YouTube URL extractor discovery to solve mobile platforms' lack of video screenshot capabilities. Implemented server-side FFmpeg processing providing pixel-perfect video frame capture for mobile users with sub-second latency.

+ +

Project Duration:

+

2024 (Principal AI Engineer at Slid)

+ +

Key Technical Achievements:

+
    +
  • Mobile Innovation: Solved fundamental limitation of mobile browsers lacking video screenshot APIs
    • +
  • Server Architecture: FastAPI backend with millisecond-accurate timestamp seeking
    • +
  • FFmpeg Excellence: Pixel-perfect frame extraction with format optimization
    • +
  • Performance: Sub-second screenshot generation with global CDN delivery
    • +
  • Scale: Handling thousands of concurrent screenshot requests
    • +
  • Mobile Optimization: Adaptive image sizing and format selection for devices
    • +
+ +

Technical Implementation:

+
    +
  • Backend Stack: Python FastAPI, FFmpeg, Pillow for image processing
    • +
  • Video Processing: Frame-accurate seeking, multi-format support, quality optimization
    • +
  • Storage: AWS S3 with CloudFront CDN for global distribution
    • +
  • Queue System: Redis-based job queue for asynchronous processing
    • +
  • API Design: RESTful endpoints with WebSocket support for real-time updates
    • +
+ +

Processing Pipeline:

+
    +
  • Request Handling: Video URL validation and timestamp parsing
    • +
  • Frame Extraction: FFmpeg seeking with millisecond precision
    • +
  • Image Processing: Format conversion, compression, watermarking
    • +
  • Delivery: CDN upload with signed URLs for secure access
    • +
  • Cleanup: Automatic temporary file management
    • +
+ +

Business Impact:

+
    +
  • Mobile User Enablement: Extended full platform capabilities to mobile users
    • +
  • User Satisfaction: 95% satisfaction rate for mobile screenshot feature
    • +
  • Market Expansion: Opened mobile-first markets previously inaccessible
    • +
  • Revenue Impact: 30% increase in mobile user subscriptions
    • +
+ +

Technical Innovations:

+
    +
  • Timestamp Precision: Sub-frame accuracy using FFmpeg's advanced seeking
    • +
  • Adaptive Processing: Dynamic quality based on device capabilities
    • +
  • Caching Strategy: Multi-layer caching reducing repeated processing
    • +
  • Error Recovery: Graceful degradation with fallback mechanisms
    • +
+ +

Integration Features:

+
    +
  • Cross-Platform Support: iOS, Android, mobile web compatibility
    • +
  • API Integration: Seamless integration with existing mobile apps
    • +
  • Batch Processing: Multiple screenshots in single request
    • +
  • Real-time Updates: WebSocket notifications for processing status
    • +
+ +

Skills Demonstrated:

+

Video Processing, FFmpeg, FastAPI, Cloud Architecture, Mobile Development, Performance Optimization, System Design

+
\ No newline at end of file diff --git a/_projects/9_project.md b/_projects/9_project.md new file mode 100644 index 0000000..34bfef9 --- /dev/null +++ b/_projects/9_project.md @@ -0,0 +1,79 @@ +--- +layout: page +title: Multi-Agent Transcript Correction +description: LangGraph-based agent system for AI dubbing pipeline transcript correction +img: +importance: 9 +category: work +--- + +
+

Description:

+

Built sophisticated multi-agent system for AI dubbing pipeline transcript correction using LangGraph. Developed specialized correction agents with custom reasoning tools, confidence scoring, and validation mechanisms that seamlessly integrated with existing dubbing workflows while significantly improving quality metrics.

+ +

Project Duration:

+

2024 (Principal AI Engineer at Slid/Bebridge)

+ +

Key Technical Achievements:

+
    +
  • Agent Architecture: LangGraph-based orchestration with specialized correction agents
    • +
  • Custom Reasoning Tools: Built domain-specific tools for transcript validation
    • +
  • Confidence Scoring: Multi-stage confidence assessment for quality assurance
    • +
  • Multi-Modal Processing: Integration of audio analysis with text processing
    • +
  • Production Integration: Seamless workflow integration without disruption
    • +
  • Quality Improvement: 40% reduction in transcript errors
    • +
+ +

Agent System Architecture:

+
    +
  • Orchestrator Agent: Main coordinator managing sub-agent workflows
    • +
  • Language Detection Agent: Identifies source language and dialect variations
    • +
  • Context Analysis Agent: Understands domain-specific terminology and context
    • +
  • Correction Agent: Applies corrections with reasoning explanations
    • +
  • Validation Agent: Final quality check with confidence scoring
    • +
+ +

Technical Implementation:

+
    +
  • Framework: LangGraph for agent orchestration, LangChain for tool integration
    • +
  • LLM Integration: Google Gemini 2.5 Flash for fast processing
    • +
  • Custom Tools: Audio analysis, terminology databases, correction rules
    • +
  • State Management: Persistent agent memory across correction sessions
    • +
  • Monitoring: Real-time agent performance tracking and debugging
    • +
+ +

Reasoning & Decision Making:

+
    +
  • Chain of Thought: Explicit reasoning paths for each correction
    • +
  • Confidence Metrics: Probability scores for suggested corrections
    • +
  • Fallback Logic: Human-in-the-loop for low-confidence segments
    • +
  • Learning Loop: Feedback incorporation for continuous improvement
    • +
+ +

Business Impact:

+
    +
  • Quality Metrics: 40% reduction in post-production corrections needed
    • +
  • Processing Speed: 3x faster than manual correction workflows
    • +
  • Cost Reduction: 50% decrease in human review requirements
    • +
  • Scalability: Enabled handling of 10x more dubbing projects
    • +
+ +

Technical Innovations:

+
    +
  • Domain Adaptation: Custom fine-tuning for industry-specific terminology
    • +
  • Multi-Agent Consensus: Voting mechanism for high-stakes corrections
    • +
  • Explainable Corrections: Each change includes reasoning explanation
    • +
  • Adaptive Processing: Dynamic agent selection based on content type
    • +
+ +

Integration Features:

+
    +
  • API Design: RESTful endpoints for dubbing pipeline integration
    • +
  • Batch Processing: Efficient handling of multiple transcripts
    • +
  • Version Control: Tracking of all corrections with rollback capability
    • +
  • Export Formats: SRT, VTT, and custom dubbing formats support
    • +
+ +

Skills Demonstrated:

+

AI Agent Development, LangGraph, LangChain, Multi-Agent Systems, NLP, Production ML, System Integration, API Design

+
\ No newline at end of file diff --git a/docker-compose.yml b/docker-compose.yml index e6068db..6f5d092 100644 --- a/docker-compose.yml +++ b/docker-compose.yml @@ -3,7 +3,6 @@ version: "3" services: jekyll: image: amirpourmand/al-folio:latest - build: . ports: - 8080:8080 volumes: From 6caceabff05d3900af71f31c09debd50767afe27 Mon Sep 17 00:00:00 2001 From: Hamza Date: Mon, 4 Aug 2025 11:26:07 +0300 Subject: [PATCH 2/4] Update X handle, enable announcements, and adjust Instagram ID in site configuration. --- _config.yml | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/_config.yml b/_config.yml index 977bbf5..99bb411 100644 --- a/_config.yml +++ b/_config.yml @@ -67,7 +67,7 @@ og_image: # The site-wide (default for all links) Open Graph preview image github_username: untilhamza # your GitHub user name gitlab_username: # your GitLab user name -x_username: untilhamza # your X handle +x_username: hamza_kyamanywa # your X handle mastodon_username: # your mastodon instance+username in the format instance.tld/@username linkedin_username: untilhamza # your LinkedIn user name telegram_username: # your Telegram user name @@ -94,7 +94,7 @@ lastfm_id: # your lastfm id spotify_id: #31zhkmixlf3hxgj3c23tlvr7neya # your spotify id pinterest_id: # your pinterest id unsplash_id: # your unsplash id -instagram_id: #untilhamza # your instagram id +instagram_id: untilhamza # your instagram id facebook_id: # your facebook id youtube_id: # your youtube channel id (youtube.com/@) discord_id: #836885952219447306 # your discord id (18-digit unique numerical identifier) @@ -172,7 +172,7 @@ collections: permalink: /projects/:path/ announcements: - enabled: false + enabled: true scrollable: true # adds a vertical scroll bar if there are more than 3 news items limit: 5 # leave blank to include all the news in the `_news` folder From 409f088c419bed77a89e12b1c5039f90e4b41ab1 Mon Sep 17 00:00:00 2001 From: Hamza Date: Mon, 4 Aug 2025 11:30:01 +0300 Subject: [PATCH 3/4] Disable announcements to fix deployment error Disabled announcements section by setting enabled: false in _config.yml to resolve GitHub Actions deployment failure. --- _config.yml | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/_config.yml b/_config.yml index 99bb411..89c61dd 100644 --- a/_config.yml +++ b/_config.yml @@ -172,7 +172,7 @@ collections: permalink: /projects/:path/ announcements: - enabled: true + enabled: false scrollable: true # adds a vertical scroll bar if there are more than 3 news items limit: 5 # leave blank to include all the news in the `_news` folder From eb9c2934efcd36b493e70aea3c233722eba50b02 Mon Sep 17 00:00:00 2001 From: Hamza Date: Mon, 4 Aug 2025 11:32:42 +0300 Subject: [PATCH 4/4] Fix profile image file extension case mismatch Changed prof_pic_2.JPG to prof_pic_2.jpg in about.md to match actual filename and resolve deployment error. --- _pages/about.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/_pages/about.md b/_pages/about.md index d6ccc9d..8f4e670 100644 --- a/_pages/about.md +++ b/_pages/about.md @@ -6,7 +6,7 @@ subtitle: #Affiliations. Address. Contacts. Moto. Etc. profile: align: right - image: prof_pic_2.JPG + image: prof_pic_2.jpg image_circular: false # crops the image to make it circular more_info: >

Technical Founder @ HoverNotes