ML/AI Engineer with a knack for research. Five years building production systems, now focused on applied machine learning and agentic AI. I gravitate toward projects that require understanding a system deeply enough to build it from the ground up, rather than relying on existing APIs.
- Agentic LLM systems — designing multi-step reasoning pipelines that convert unstructured input into structured, verifiable output
- ML fundamentals — implemented a GPT from scratch (attention, transformers, KV-caching, custom tokenizer) to build a thorough understanding of the underlying architecture
- Computer vision — object detection (YOLOv8) applied to real-world document and image analysis tasks
| Project | Description |
|---|---|
| agentic-trial-matching | Converts free-form clinical trial eligibility criteria into structured feasibility queries, using LLM-driven concept resolution, clinical code mapping (ICD-10, SNOMED, LOINC), and automated SQL generation |
| neetcode-gpt | A GPT implemented from scratch in PyTorch, including multi-head attention, grouped query attention, KV-caching, a custom BPE tokenizer, and a complete training pipeline |
| sigcore | Signature detection in documents using YOLOv8 |