Software Engineering Student · Building full-stack applications, ML pipelines, and data-driven solutions.
I'm a Computer Science / Software Engineering student at the University of Georgia with hands-on experience shipping production-grade web apps, deep learning systems, and enterprise data pipelines. I love tackling hard problems — from training attention-based neural networks under extreme noise to automating 45K+ file workflows for real business operations.
| Category | Technologies |
|---|---|
| Languages | Python · JavaScript · TypeScript · SQL · HTML · CSS · YAML |
| Frontend | React · Next.js · Vite · Tailwind CSS · Framer Motion · Radix UI |
| Backend | FastAPI · Django REST Framework · Flask · Node.js · Gunicorn |
| AI / ML | PyTorch · scikit-learn · OpenAI Whisper · Hugging Face Transformers · Google Gemini API |
| Databases | PostgreSQL (Supabase) · SQLite |
| Data & Analytics | Pandas · Power BI · Azure Maps · Firecrawl · Playwright |
| DevOps & Tools | Docker · GitHub Actions · Vercel · Render · Railway · Git |
- Engineered a 31.5M-parameter Attention U-Net with CBAM and anti-alias blur-pool downsampling, achieving a validation Dice score of 0.886 and IoU of 0.796 on segmenting organic shapes under extreme variable noise (SNR from >30 dB to <0 dB).
- Implemented a 4-phase curriculum learning strategy (Easy → Extreme noise) with custom composite loss (Dice + BCE + Boundary Tversky), training 80 epochs on an RTX 4070 Ti SUPER with mixed-precision (fp16).
- Built a full deployment pipeline — FastAPI backend on Railway, glassmorphism web dashboard on Vercel, and 379 MB model checkpoint hosted on Hugging Face — featuring interactive confidence heatmap overlays and dual-pane comparison viewers.
- Stack: PyTorch · FastAPI · OpenCV · Docker · Vercel · Railway · Hugging Face
- View Project
- Architected a full-stack cinema booking system with Next.js 16 / React 19 frontend and Django REST Framework backend, enabling end-to-end movie browsing, showtime selection, seat reservation, and PCI-aware payment processing.
- Designed a real-time seat availability system with showroom management, supporting concurrent bookings across multiple theaters and showtimes with Supabase PostgreSQL as the persistence layer.
- Developed a comprehensive user experience including authentication, favorites, reviews, profile management, and an admin panel for movie/showroom CRUD operations, with Resend-powered transactional emails for password resets.
- Stack: Next.js · React 19 · Django REST · Tailwind CSS · Supabase (PostgreSQL) · Vercel · Render
- View Project · Live Demo
- Built a wizarding-themed bank statement analyzer that parses uploaded PDF statements, classifies transactions into 10 spending categories using a Naive Bayes ML classifier (TF-IDF + Multinomial NB), and delivers AI-generated financial advice via Google Gemini with ElevenLabs text-to-speech narration.
- Integrated Supabase Auth (PKCE + Row Level Security) with a React 18 frontend featuring drag-and-drop upload, gamified wizard-rank progression (Muggle → Order of Merlin), and a mobile-first responsive design with Framer Motion animations.
- Deployed on Vercel (frontend) and Render (backend) with a FastAPI service layer handling PDF extraction via pdfplumber, ML inference, and Gemini API orchestration.
- Stack: React · FastAPI · scikit-learn · Google Gemini · Supabase · ElevenLabs · Vercel
- View Project · Live Demo
- Developed an end-to-end audio intelligence pipeline that transcribes uploaded audio files using OpenAI Whisper, summarizes transcripts with Google T5 via Hugging Face Transformers, and plays back summaries through Google Cloud Text-to-Speech — all in a single unified processing endpoint.
- Collaborated on a 4-person team, owning UI/UX design and frontend integration with a React 19 / Vite stack and Flask 3.0 backend serving ML inference and TTS endpoints.
- Stack: React 19 · Flask · OpenAI Whisper · Hugging Face T5 · Google Cloud TTS · SQLite
- View Project
- Scraped and analyzed U.S. air rifle retailer data (Pyramyd Air, Airgun Depot) using Firecrawl and Playwright to shift Norica's market approach from "Product Mix" to a data-backed "Sales Mix" strategy, identifying competitor gaps in the Recoil-Free segment.
- Visualized feature-to-sales correlations and regional profitability scores in Power BI dashboards with Azure Maps geospatial heat maps, using review volume/velocity and best-seller rankings as sales proxies.
- Stack: Python · Pandas · Firecrawl · Playwright · Power BI · Azure Maps · Supabase (PostgreSQL)
- View Project
- Automated a 7-step data processing pipeline that scans, extracts, and reconciles Request for Proposal (RFP) data across 45,231 file paths, filtering down to 8,742 relevant files using fuzzy string matching (fuzzywuzzy + Levenshtein distance) against CRM account databases.
- Engineered an intelligent "failed twice" exclusion system that permanently flags unreliable entries, combined with centralized configuration, automatic backups, multi-encoding support (UTF-8/UTF-16), and batch-orchestrated progress tracking.
- Stack: Python · Pandas · fuzzywuzzy · openpyxl · Windows Batch scripting
- View Project
- Designed a modern, responsive developer portfolio featuring dark/light theme toggling, animated starfield backgrounds, scroll-reveal animations, and section-based navigation — built with React 19, Vite 7, and Tailwind CSS 4.
- Stack: React 19 · Vite · Tailwind CSS · Radix UI · Lucide React
- View Project
- Exploring advanced deep learning architectures and deployment at scale
- Deepening expertise in cloud infrastructure, CI/CD, and DevOps workflows
- Building more full-stack applications with Next.js and modern React patterns
- LinkedIn: linkedin.com/in/rmeyer3
- Portfolio: ryanmeyer.vercel.app
- Email: ryanpaulmeyer@gmail.com