AI Research Analyst at Furnitureland South (the world's largest furniture store) and CS graduate from UNC Charlotte. I build AI systems that actually get used—not just demos, but production platforms handling real users and real scale.
My work spans the full stack of AI engineering: from designing AWS serverless architectures to building conversational AI with Microsoft Copilot Studio to optimizing vector search pipelines that turned a 15-second search into a sub-500ms experience.
|
Visual similarity search across 120K+ furniture products. Upload an image → get instant matches based on what it looks like, not just keywords. The problem: Sales reps needed to find "that sofa the customer saw somewhere" in a massive catalog. The solution: CLIP embeddings + FAISS vector search + optimized inference = magic. |
|
Technical highlights:
- CLIP-based dual image/text embeddings
- FAISS with optimized index structures
- Evaluated Milvus for filtered metadata queries
- Dockerized microservices architecture
- FastAPI backend, React frontend
|
AI assistant for sales reps built on Microsoft Copilot Studio, integrated with NetSuite ERP for live vendor data, promotions, pricing, and customer history. Deployed to production — used daily by the sales team. |
|
Technical highlights:
- Microsoft Copilot Studio topic/flow architecture
- Boomi API integration for live NetSuite data
- SuiteQL queries for complex vendor lookups
- Analytics pipeline for usage insights
- Designed data sync automation (replaced manual CSV uploads)
|
Designed and prototyped an automated call transcription system for Zendesk support tickets using OpenAI Whisper on AWS. Built local backfill tooling that processed 3-6 months of historical call recordings. |
|
Technical highlights:
- Benchmarked
faster-whispermodels (tiny → medium) for cost/accuracy tradeoffs - Compared against Azure Speech Services
- Built local Python app for Zendesk → Whisper → transcript attachment pipeline
- Documented cost projections and ECS sizing recommendations
After-hours conversational AI on the company website. Handles FAQs, guides users through processes, and intelligently routes to human agents.
- Knowledge base article creation for coverage gaps
- Conversational flow design for complex tasks
- Intent refinement and entity recognition improvements
- QA testing with real inquiry data
📋 Full Breakdown
| Category | Technologies |
|---|---|
| Languages | Python, TypeScript, JavaScript, SQL (SuiteQL), Java |
| AI/ML | CLIP, FAISS, Whisper, Embedding pipelines, Vector search |
| Backend | FastAPI, Node.js, Spring, RESTful APIs |
| Frontend | React, Next.js, HTML/CSS |
| Cloud & Infra | AWS (Lambda, S3, SQS, ECS Fargate, API Gateway), Docker |
| Integrations | NetSuite, Zendesk, Microsoft Copilot Studio, Boomi |
| Tools | Git, Jira, VS Code |
| Project | Description |
|---|---|
| Resume & Pitch Agent | AI app that optimizes resumes for specific jobs and generates pitch decks |
| ARC-AGI Benchmarking | Testing LLM baselines on the ARC-AGI reasoning benchmark |
| Algo Vault | Polyglot algorithm implementations (Go, TS, Python, JS, C#, C++) |
current_work = {
"🔍 Search & Retrieval": "Vector databases, semantic search, sub-second latency",
"🤖 Conversational AI": "Enterprise chatbots that integrate with real systems",
"☁️ Cloud Architecture": "Serverless pipelines, cost-optimized ML inference",
"🔗 System Integration": "Making AI work with NetSuite, Zendesk, and legacy systems",
"📊 AI Analytics": "Understanding how AI tools are actually being used"
}B.S. Computer Science — UNC Charlotte (Dec 2026)
I'm always interested in discussing AI systems that ship, search infrastructure, or interesting engineering challenges.
"The best AI is the AI that gets used."