11th grade · NES International School, Mumbai · building things that matter
I'm a high school developer who builds AI-driven projects that solve real problems — not demos that gather dust. Right now I'm most interested in algorithmic fairness, ML systems, and building tools that make complex ideas legible to anyone.
I learn by shipping. If something breaks, I want to know why it broke — not just how to patch it.
Fair Code — An open-source research project exposing and fixing bias in real-world AI systems. Audits four deployed algorithms — criminal justice (COMPAS), hiring, lending, and healthcare — using real datasets. Every experiment follows the same pipeline: train a biased model, measure the fairness gap, strip protected attributes and proxy variables, retrain, measure again. Up to 97.3% bias reduction achieved. Includes six deep-dive explainers on proxy variables, sampling bias, SHAP values, equalized odds, disparate impact, and why fairness metrics conflict.
Python · scikit-learn · pandas · SHAP · Fairlearn
↗ Live · ↗ GitHub · ⭐ 27 · 🍴 8
CardioAI — A cardiovascular risk predictor that turns clinical health parameters into an instant ML-based risk assessment. Users enter vitals like blood pressure, cholesterol, glucose, and BMI; the model outputs risk scores with an optional hypertension assessment module. Includes an AI assistant, interactive data visualisations, and a risk-reduction guide.
Python · ML · React · Next.js
↗ Live · ↗ GitHub · ⭐ 4
Daily drivers: Python, JavaScript, React, HTML/CSS, Git
Have shipped with: scikit-learn, pandas, Next.js, TailwindCSS, Flask
Currently exploring: ML fairness, agentic AI systems, systems-level thinking, Rust
I don't build things to put them on a resume. I build them because the problem is interesting and the solution isn't obvious yet. I'm skeptical of hype — including my own — and I'd rather show measurable results than make claims.
Fair Code exists because I wanted to understand whether AI bias was actually fixable or just theoretically fixable. Turns out: it's fixable. The code is public. The numbers speak for themselves.
