I build Python AI applications, especially LLM systems that work with documents and tools. I am looking for AI Engineer and Software Developer roles where I can turn model ideas into reliable software.

I began with AnyDoc, a Streamlit app that chunks PDFs or text and uses Groq’s Llama 3.1 API for summaries and Q&A. Live demo.

Then I built AutoWebsite-Maker: Gemini and Playwright adapt a landing page’s headlines and CTAs to an ad, with a rule-based fallback. Live demo.

For computer vision, ACS-SegNet runs a trained H&E segmentation checkpoint through FastAPI and returns a mask, heatmap, and overlay. Demo assets.

More recently, I built Shortfall Guard, an agentic cash-flow prototype with human approval, a deterministic fallback, and six automated tests.

I also contribute upstream. In AiiDA, I have two merged pull requests:

• PR #7400 fixes multiline PBS qstat parsing and resolves #7395.
• PR #7385 makes remote cleanup warn and continue when AuthInfo is missing.

LLM: Gemini, Groq/Llama, prompting, tool calling, LangChain, LangGraph, and document processing.

Delivery: Python, FastAPI, Docker, pytest, PostgreSQL, Redis, and React.

ML: PyTorch, CNNs, Transformers, U-Net, XGBoost, Random Forest, and SHAP.

I am now learning production RAG: vector retrieval, grounded answers, and LLM evaluation. Useful AI needs reliable context, visible failure modes, and a way to measure whether an answer helps.

Portfolio · Email