Tools change every year. The way of working stays.
Understand the problem, build the smallest honest version, and put it in front of a real person. These are the tools I currently reach for.
LLMs and retrieval-grounded systems, deep learning, transfer learning, computer vision, and NLP. I care most about the part where a model leaves the notebook and survives contact with real users.
PyTorch, TensorFlow, Hugging Face, LangChain, scikit-learn, OpenCV
The products I build and run need full stacks behind them, so I work across the whole thing: APIs, databases, mobile, and deployment.
Python, TypeScript, FastAPI, NestJS, React, Next.js, Flutter, PostgreSQL, MongoDB, Supabase, Docker
Statistical analysis and time series work, mostly in service of decisions rather than dashboards.
Python, pandas, statistical modeling, time series analysis