End-to-end ML pipeline predicting insurance costs using features like BMI, smoking status and region. GradientBoosting model (R² 0.866) (not very good as the data size was small 1338 samples), FastAPI backend, Streamlit frontend, containerized and pushed to Docker Hub.

A Streamlit tool that sends the same coding task to 8 LLMs together and evaluates each response. Python code are checked with test-cases and Lean 4 code are checked by the compiler and inspected for incomplete proofs. An iterative feedback loop lets users re-prompt each model through up to 5 refinement. All responses, timings, and scores are logged to SQLite for analysis

A paid course platform with 100 daily lessons (currently only for Python), in-browser code execution using Piston API, auto graded problems, AI hints, and progress tracking with streaks. Built everything solo, course content, full-stack webapp, payments and deploy.