Projects
Research and engineering across machine learning, causal inference, computer vision, poker AI, and trading.
Pro Valorant: Match Outcomes & Kill Props
Predictive models on 28K pro Valorant matches (600K player-maps) — GBM beats Elo on match outcomes (62.4% vs 60.9%), a maps-1-and-2 series-kills model beats the rolling-mean baseline by 9.6% (>36σ), and a Map-3 decider model lifts another 3.4%. With honest CV and zero-edge sanity checks.
Subscription Churn Prediction & Causal Uplift Modeling
KKBox subscription churn pipeline (real Kaggle data, 6.8M members) — LightGBM at ROC AUC 0.866, then a causal uplift model that retains +92% more value than risk-targeting on the same budget. Engagement-decay naive estimate flips sign under adjustment.
Crypto Strategy Discovery: Robust BTC & ETH Research
A research project that hunts for genuinely profitable, non-overfit BTC/ETH strategies. Frozen 30% out-of-sample holdout, walk-forward parameter selection, deflated Sharpe, parameter-neighborhood robustness, realistic frictions. Returns are easy to fabricate; statistical honesty is hard.
Deep CFR for 5-Card PLO Heads-Up
Stage 6 of a neural-CFR research arc — porting Deep CFR from HUNL to 5-card PLO heads-up. Composition-dependent encoder, opp-value board cache (99.4% hit rate, 58× steady-state encode speedup), equity-pretrained warm start, and an honest profile of why scaling the action grid is harder than scaling the cards.
Deep CFR for Heads-Up No-Limit Hold'em
A full neural CFR pipeline for HUNL — game logic, feature encoder, V/R/S networks, batched sigma scheduler, exploitability evaluation. 200 iterations × K=10,000 traversals at seed 42, 17.86h wall, no NaN/Inf. Stage 4 of a six-stage research arc culminating in PLO5.
Polymarket Research Toolkit
Research-first scraping + walk-forward backtester for Polymarket. Six strategies, deflated Sharpe ratios, and conservative cost models. Designed to fail loudly when no real edge exists — and it does.
Hierarchical Bayesian Skill Rating for Pro Valorant
Hierarchical Bradley–Terry with map, region, and time pooling. On 86 held-out cross-regional matches, log loss 0.600 vs Elo's 0.640 (95% CI [−0.019, +0.098]); ECE drops 0.092 → 0.060 and confident calls (≥70%) land at 78% empirical.
Causal Effect of the Chamber Nerfs on Pro Valorant
Interrupted time series + difference-in-differences across maps to quantify the meta restructuring caused by Patch 5.12. Chamber pick rate collapsed −78pp at the patch (p < 1e-27); sentinel-pick entropy rose +0.42 bits; placebo dates and bandwidth sweeps both check out.
Switchback Experiments on a Simulated Marketplace
Built a two-sided rideshare marketplace, broke per-rider A/B testing on it (208% biased), and recovered the true effect with a switchback design. Worked-through bias-variance tradeoff in window length, cluster-robust SEs, and a power analysis that explains why marketplaces need 6–8 week experiments.
Pickleball Vision: CV-Driven Match Analytics
A computer-vision pipeline that turns a fixed-camera pickleball clip into an annotated video with player tracking, ball tracking, court geometry, top-down minimap, and per-frame analytics — ball speed, player speed/distance, shot count. YOLOv8 + a fine-tuned ResNet50 court keypoint regressor + iterative homography refinement.
LSTM-Driven Poker Analytics & Bluff Prediction Platform
Developed an end-to-end machine learning pipeline that extracts, cleans, and feature-engineers over 5.7k real-money hand histories from PokerNow.club. Leveraging advanced feature engineering (bet ratios, decision times, board evaluations with Ace detection, and dynamic positional metrics) and a custom LSTM model with dynamic bucketing, the system predicts bluff versus value betting with a test AUC of 0.77. Hyperparameter tuning, cross-validation, and class balancing were key to optimizing performance.
LEAP Trading Strategy: Leveraged Long-Dated Options Backtest
A research project on whether systematic LEAP (long-dated deep-ITM call) strategies can beat buy-and-hold on a risk-adjusted basis. Self-funded variants, drawdown stability, drip-DCA sweeps over moneyness × tenor, and a real backtest on historical option chain data.
ORB Algorithmic Day-Trading System
Opening-Range-Breakout strategy on TQQQ with an XGBoost gating layer that decides whether to take the day's signal at all. SQL-driven feature pipeline, multi-interval simulation, and a +19.1% annualized return improvement over the unfiltered ORB baseline.
CNN-Based Age Prediction System
Trained a CNN-based age prediction model using PyTorch, the UTK dataset, and a ResNet10 architecture to predict ages with an average error of ±4 years—optimized via hyperparameter tuning and deployed with a Streamlit UI for real-time inference.
No-Bust 21st Century Blackjack — Monte Carlo + CDZ⁻ Solver
A multi-process Python simulator with a Tkinter GUI for the California card-room variant *No Bust 21st Century Blackjack*. Built around exact composition-dependent (CDZ⁻) strategy solving plus Numba-JIT'd hand play. Includes the "no-bust comparison" rule, surrender at any decision point, split-aces special handling, DAS, and the buster side bet.
Ethereum Smart Contract for NFT Generation & Minting
Designed and deployed an Ethereum smart contract for NFT generation and minting, supporting 10,000+ unique ERC-721 assets with an optimized Python image pipeline, reduced gas fees by 15%, and a user-friendly UI that facilitated over 1,000 transactions.