I build production-grade AI systems end-to-end — from model training & optimization to inference APIs, mobile apps, and cloud deployment.
- B.A. Computer Science & Mathematics — Bennington College (May 2025)
- Focus: Applied ML, AI Infrastructure, Full-Stack Engineering
- Interests: LLM systems (RAG/agents), multimodal AI, speech/vision, mobile-first AI products, scalable platforms
- LLM-powered products: RAG, agents, tool-calling, embeddings, evals
- Real-time inference APIs: streaming speech/text, low-latency pipelines
- Mobile + web apps: AI-native UX, voice/chat/search flows
- Optimization: latency/cost tuning, batching, quantization, GPU/CPU tradeoffs
- Deployment: serverless, containers, GPU-backed services, CI/CD
Languages
ML / AI
LLM Systems / Data
Frontend / Mobile
Cloud / DevOps
Production-ready voice AI system combining:
- Whisper ASR + speaker diarization + speaker verification (embeddings)
- Streaming + batch inference, real-time APIs for web/mobile clients
- Containerized GPU deployment with autoscaling
- Lower cost + faster inference via quantization & mixed precision
I built a spec-driven rendering engine that turns an Excel-based schema into a dynamic, validated UI and exports PDF + MISMO/XML.
What it does
- Parses an Excel spec (UID/xPath bindings, containers, enums/formats, cardinality, rules)
- Generates a Section Tree + Field Registry to drive the UI
- Uses a Rule Engine (required/visible/validate) enforced at runtime
- Centralized XMLStore as the “source of truth”
- React Context + hooks for read/write bindings across inputs
- PDF renderer mirrors the same UI state
- XML builder exports MISMO nodes while honoring R/CR + repeatable sections
- Built dermatology models fair across Fitzpatrick I–VI
- Used group-aware sampling, reweighted losses, calibration + ensembling
- Placed 2nd / 300+ teams