I turn dataDataPandas, NumPy, and SQL, plus engineered features like Sharpe, Sortino, and drawdown. into workingWorkingShipped, not just notebooks: FastAPI backends and React dashboards on Render and Vercel, and Streamlit apps. AIAIPPO reinforcement learning, DistilBERT and FinBERT sentiment, and RAG with Gemini..

A PPO-based portfolio manager with a custom Gymnasium environment that shifts allocation across the S&P 500, NASDAQ, Gold, Oil, and NIFTY 50. It uses monthly return, volatility, Sharpe, Sortino, drawdown, and correlation features.

On 2023–2025 backtesting it returned 66.4% in total, 19.9% CAGR, and a 1.87 Sharpe ratio. I also ran FinBERT sentiment experiments and built FastAPI endpoints behind a React and TypeScript dashboard, deployed on Render and Vercel. I’m still adding more features.

Source on GitHub

A machine learning system that predicts cyclone trajectories over the North Indian Ocean, covering the Bay of Bengal and the Arabian Sea.

It learns from historical cyclone data to forecast future storm positions using Random Forest and XGBoost models. An OpenStreetMap view shows the predicted paths, and a FastAPI backend serves the forecasts.

Source on GitHub

A customer sentiment platform for text and CSV reviews. It runs a Transformer model over the reviews and shows the results in interactive charts so you can explore what customers are saying.

Source on GitHub

A movie recommendation system built on the Kaggle TMDB 5000 dataset. Pick a movie and the app suggests similar ones, using a Streamlit interface over a Scikit-learn model.

Source on GitHub