Work
These are my own public projects. None is client work, and each is tagged for what it is. If a project is not listed here, it is not because it does not exist, but because it is not mine to show.
Streamlit app that cleans an uploaded CSV, lets you pick a target and a simple model, and visualises the result in 2D and 3D. It is built for exploring data and understanding models, not for producing predictions.
Streamlit app for a Data Science course. It profiles an uploaded CSV, checks for missing values and duplicates, detects whether the task looks like classification or regression, gives a readiness score and answers common next-step questions.
A small test of one question: does a lightweight verification loop (source trust tiers plus syntax checks) reduce confidently wrong answers in a self-teaching local agent? It runs on a 1.5B-parameter model with no GPU. On 8 developer questions, the naive version labelled all 6 blog-sourced answers as trusted and the verified version labelled none. That is a small sample, and the report lists its limitations.
Send a sample or a description. I will tell you what I would do and what I would need.