Designed and built an end-to-end autonomous research platform that automates the scientific discovery process for cancer biology. Developed a multi-stage pipeline that scrapes PubMed/arXiv for literature, constructs dynamic knowledge graphs using NetworkX, and generates novel gene-mechanism hypotheses via Groq-powered LLMs. Engineered a high-fidelity biological simulation (ODE-based Gymnasium environment) where a PPO-based Reinforcement Learning agent performs virtual experiments to identify causal regulatory interactions. Features a premium 3D virtual laboratory with glassmorphism UI, real-time causal graph visualization, and automated technical discovery reports.

Built an end-to-end quantum-aware security platform that analyzes IoT data metadata to predict cryptographic risk and generate cost-optimized migration strategies for classical, hybrid, and post-quantum cryptography. Implemented ML-based risk classification, policy-driven migration planning, and real cryptographic benchmarking of RSA, ECC, and Kyber. Designed a Flask REST API with CSV ingestion, explainable risk reports, and exportable migration roadmaps, enabling enterprises to plan secure and efficient transitions to post-quantum cryptography.

A secure, decentralized voting protocol built on the Solana blockchain through transparent proposal and voting systems backed by an on-chain treasury. Features token-based voting, automated treasury management, and a high-trust dark-mode UI.

Built a full-stack quantitative finance platform that prices options in real-time using Black-Scholes analytics and Monte Carlo simulation (antithetic variates, 30k+ paths) with live market data. Engineered a discrete delta-hedging simulator that compares daily, weekly, and monthly rebalancing across 400+ simulated paths — directly quantifying the continuous-time assumption breakdown in Black-Scholes. Features an IV regime classifier, IV spike predictor, real-time mispricing detection, and a full options chain viewer with volatility smile visualization.

Developed a high-performance, production-grade CLI tool for Go that eliminates manual rebuild cycles through intelligent filesystem orchestration. Engineered a zero-dependency core utilizing fsnotify and the Go standard library, featuring a custom timer-reset debouncing algorithm to handle rapid IDE save events. Implemented robust process management using Unix process groups (PGID) to ensure clean termination of complex process trees, alongside smart crash-loop protection with exponential backoff. Optimized for scale with recursive dynamic directory watching, categorical file filtering (ignoring .git/vendor), and real-time log streaming with graceful signal handling.

Engineered a production-grade machine learning platform to predict and analyze Air Quality Index (AQI) dynamics across 291 Indian cities using a dataset of 235k records. Developed a zero-leakage ML pipeline featuring a 54-feature intelligence engine with cyclical time encodings, 30-day rolling averages, and geographic target encoding. Addressed severe class imbalance (160:1) using SMOTE to ensure high-recall detection of hazardous pollution spikes. Achieved an R² of 0.904 using a custom-tuned LightGBM ensemble. The system includes a high-performance FastAPI inference backend and a sleek, dark-mode analytics dashboard with interactive city-to-city comparisons and automated WHO-aligned health advisories.