Articles and thoughts on AI, Machine Learning, and Software Engineering
A four-page business intelligence report built on a 9-table quick-commerce dataset: a star schema model, 45+ DAX measures, and the data-quality checks that kept the numbers honest.
A two-tab Streamlit app that scores Telco customers for churn risk and recommends a specific retention action for each one. What it does, how the model was chosen, and where it stands today.
A dashboard that turns five years of official air-quality data from 11 Indian cities into something you can actually explore — plus an 'Ask AI' assistant that answers questions by calling real data tools (MCP) instead of guessing. What I built, why the raw data needed fixing first, and how the AI layer works.
Notes on teaching a 3.8B model to write SQL with QLoRA, quantizing it to run without a GPU, what the evaluation actually showed, and the four things that broke getting it onto a free server.