I'm an AI/software developer focused on LLM systems, model evaluation, retrieval, and the engineering needed to make those systems reliable.
I work mainly in Python and TypeScript, with recent projects spanning LLM fine-tuning and evaluation, RAG/GraphRAG, knowledge graphs, data curation, and full-stack software development.
I'm currently open to remote AI and software engineering roles.
A reproducible supervised fine-tuning experiment using Qwen3-4B + QLoRA for GitHub issue classification.
- Trained on 31,876 issues using repository-held-out splits
- Evaluated on a frozen 9,708-example TEST set
- Compared zero-shot, few-shot, and fine-tuned conditions
- Base zero-shot achieved 0.619 macro-F1 vs. 0.575 for the fine-tuned model
- Preserved and documented the negative result rather than tuning against TEST
- Diagnosed and corrected an evaluation-harness truncation bug before final TEST evaluation
- Used paired bootstrap confidence intervals and McNemar testing for the final comparison
A live platform for maintaining structured, versioned evidence records around important claims in AI, science, and technology.
- Designed the product direction, taxonomy, data model, and research workflow
- Built deterministic data releases with immutable publication snapshots
- Uses automated type checking, unit tests, static validation, Playwright, and Axe
- Deployed through GitHub Actions to Cloudflare Pages
Live: vydex.pages.dev
A multi-iteration project exploring retrieval for large fictional-world corpora where relevant context is not always the text most semantically similar to the current query.
Across its prototypes, VySol has explored:
- Document ingestion and configurable chunking
- Embeddings and vector retrieval
- Knowledge-graph extraction
- Entity resolution
- Graph-backed retrieval
- Source-level graph provenance
- Context inspection and failed-extraction repair
- Python/FastAPI and TypeScript/React application development
Languages: Python, TypeScript
AI / ML: LLMs, QLoRA, PyTorch, model evaluation, NLP, RAG, GraphRAG, knowledge graphs
Software: FastAPI, React, Astro, Git, automated testing, JSON
I'm interested in work involving AI engineering, LLM fine-tuning and evaluation, AI systems, retrieval/RAG, Python/backend engineering, data curation, and software engineering.