Projects
A collection of trading systems, ML models, and product experiments I've built.
Featured
Weather Prediction Trading System
Built a fully automated trading system for weather prediction markets. Implemented ensemble forecast calibration with 26% Brier score improvement via seasonal bias correction + isotonic regression, probabilistic edge detection, and Kelly criterion position sizing. Achieved 67% win rate and +14% ROI on live capital.
Lokin - Startup Automation Platform
Co-founded and built a platform that automates core startup processes for students. Developed agentic AI tools for profile scraping, teammate matching, competitor analysis, and job description generation. Led product strategy and coordinated cross-functional teams.
Semantic Search Engine
Built semantic search for 76K+ products using MiniLM embeddings and Pinecone DB. Delivers top-10 similar results with TF-IDF keywords via FastAPI/CLI interface. Achieved <300ms latency, deployed on Railway.
Competitor Intelligence API
Built competitor intelligence API for startup validation. Engineered intelligent query refinement for ambiguous product concepts, structured competitor extraction with tiered classification, and automated source/company verification. Delivers actionable pitch-ready reports in <30 seconds. Used by Lokin users and available as a standalone tool.
More Projects
Trade Analytics Interface
Designed a Python + SQL-based agentic AI interface at TP ICAP to query internal trade history. Enabled Broker/Ops self-serve analytics and automated manual retrieval, achieving 3× faster data access.
Customer Churn Prediction Model
Built a churn prediction model for a digital bank using XGBoost, achieving 74.5% AUC. Applied feature engineering and explainability techniques to drive actionable customer retention insights.
Bike-Share Demand Forecasting
Led a 4-person team to build bike-share demand forecasting using Random Forest (12% SMAPE) with lag/rolling features. Identified $300K revenue uplift potential and packaged insights into focused recommendations.
Insider Trading Anomaly Detection
Developed anomaly detection system at Abel Noser to flag potential insider trading using lead-lag analysis on stocks and linked ETFs. Part of post-trade analytics for institutional clients.