Technical Projects

Deep Learning for Conformal Mapping

• Neural networks computing Schwarz-Christoffel conformal maps, turning unit disks onto polygons
• MSE analysis with ELU activation on fixed quadrilaterals and variable-vertex fluid mechanics datasets
• Detailed exploration of ML architecture with backpropagation for gradient flow and noise reduction

Exploring Style Factors in Digital Assets

• Decomposition of crypto equity returns into systematic risk factors with academic frameworks
• Performance evaluation involving Factor Ratio metrics and advanced Momentum portfolio optimizations
• Methodological approach to isolate statistical significance of liquidity and volatility drivers

MNDY Analysis & Short Thesis

• Comprehensive fundamental analysis of Monday.com (MNDY) highlighting operational risks
• Rigorous short investment thesis developed specifically for the Point72 Academy competition finals
• Advanced discounted cash flow (DCF) valuation models and quantifiable downside risk targeting

Academic Group Submission

• Collaborative academic paper by Riyaa Randhawa, Kent Lemken, and Felipe Cardozo
• Detailed methodical breakdown involving extensive scope and structured literature review
• Presentation of empirical findings and synthesis of statistical results for portfolio optimization

Regime-Aware Multi-Agent Portfolio Allocator

• HMM regimes + LightGBM alpha across 3,500+ trading days with PPO agents in 5-phase ML pipeline
• DNN calibrates rough Bergomi volatility from 42-point IV surfaces at ms latency without numerical iteration
• Purged cross-validation with 5-day embargo; validated against MVO and equal-weight baselines

FDIC & ABS Intelligence System

• Neo4j risk engine to visualize complex financial networks and uncover hidden debt patterns
• Automated data ingestion with custom Python pipelines, streamlining disparate sources
• Graph centrality algorithms score credit exposure, sharpening detection of high-risk network links

High-Frequency Trading System

• Executes futures and options in 9ms with 24/7 market data, 0–300% leverage via IBKR data
• Pre-trade margin preview, VaR limits with auto kill-switches and persistent SQLite audit trails
• Prometheus metrics + Dockerized Azure deployment with real-time PnL and latency dashboards

Risk-Neutral Option Pricing Engine

• 6 stochastic models via Carr-Madan FFT — sub-$0.01 pricing error
• Implied volatility surface calibration via differential evolution on Heston models; <5 bps RMSE
• VaR, CVaR, and Greeks via Fourier inversion — O(N⁻²) convergence vs. Monte Carlo's O(N^−½)

Systematic Portfolio Optimization

• 4 strategies: Max Sharpe, Min Vol, Risk Parity, CVaR — achieving 1.34 Sharpe ratio
• Black-Litterman returns + Ledoit-Wolf shrinkage covariance; 12.28% annualized returns on 10-asset universe
• Walk-forward backtest with mark-to-market drift, turnover constraints, and transaction costs over 5 years

2026 FIFA World Cup Match Outcome Prediction

• 100K Monte Carlo + ML simulations; >65% match prediction accuracy
• 45+ models parameterized — <0.8 log loss, <0.15 Brier score across 153 probability indicators
• Calibrated non-stationary models on 12 past tournaments using FIFA & Elo data

Deep Learning for PM2.5 Prediction

• Logistic Regression, Gaussian Processes, and Neural Networks across 420K urban air quality records
• 74% empirical coverage for 95% prediction intervals; calibrated multi-horizon uncertainty via Bayes
• 50+ meteorological features, temporal validation, automated backtesting on Snowflake ML

Deep Learning for Conformal Mapping — Research Paper

• Neural networks compute Schwarz-Christoffel conformal maps, turning unit disks onto polygons
• 0.0038 MSE with ELU activation on fixed quadrilaterals; 0.0194 on variable-vertex fluid mechanics datasets
• MLP ReLU with Xavier initialization; backpropagation for gradient flow and noise reduction

Numerai Tournament Q3 2024 — 92nd Percentile

• ML algorithm blending 6 LightGBM regressors on Numerai v5.2 targets with rank-gaussian normalization, 156-era folds
• Feature neutralization at 0.5 proportion via pseudoinverse projection to maximize Meta Model Contribution
• Expanding-window walk-forward CV with 8-era purge gap — ranked 843 / 10,000

Exploring Style Factors in Cryptocurrencies — Research Paper

• Decomposed crypto risk returns into systematic factors with long-short Factor Ratio and Momentum portfolios
• 160.38% annualized returns, Sharpe ratio of 0.93
• Proved liquidity and volatility are statistically significant return drivers