Experience
June 2025 - Present
AI/ML Research Assistant
San Francisco State University
- Architected an agentic course equivalency system achieving 98% F1-score, automating transcript parsing and pairwise semantic matching with FERPA-compliant local PII redaction.
- Fine-tuned Sentence Transformer model (BGE) using triplet loss on 9,192 course pairs, achieving 5.3% accuracy gain on semantic similarity benchmarks.
- Quantized BGE embedding model via CTranslate2, reducing model size by 74% and achieving 2.81x inference speedup with negligible accuracy loss.
- Deployed 5 containerized ML microservices on AWS Lambda, achieving 489ms p95 warm invocation latency with independent scaling across the equivalency pipeline.
January 2024 - May 2024
Machine Learning Engineer Intern
Bluepixel Technologies
- Adapted LlaMA-3 on clinical records using PEFT/LoRA, reducing perplexity by 27% and improving domain-specific terminology alignment.
- Built a real-time patient monitoring pipeline via LangChain and InfluxDB, automating structured clinical JSON extraction for 30+ patient streams with 99%+ parse success rate.
- Implemented a stateful conversational layer over LlaMA-3 with session-scoped context management and domain-restricted guardrails, supporting safe multi-turn health queries.