Published researcher · Hackathon finalist · Building at the intersection of ML systems, security, and quantitative finance.
- PCAM — Query-conditioned precision control in frozen associative memory · Proved 90/90 benchmark structurally unreachable · Δ=+0.124 over baseline (5 seeds, zero regressions) · Manuscript in preparation
- ICCI 2026 — Published paper on cybersecurity attack surfaces in invasive Brain-Computer Interfaces · Sole student author · Year 1
- RiskEngine-Java — Real-time financial risk engine · Java 21 + Spring Boot + Kafka + ONNX Runtime + TimescaleDB · sub-50ms p99 latency · 112+ tests, CI via GitHub Actions
- God's Eye — Rust SIEM · Bayesian ensemble (XGBoost + Autoencoder + LSTM + Sigma rules) evaluated on CICIDS-2018 + DARPA · ≥88% precision / ≤5% FPR
- Multi-Asset-Portfolio-Stress-Tester — Full-stack portfolio risk engine · VaR/CVaR across 4 methodologies (historical, parametric, Monte Carlo with Cholesky-decomposed Student-t, macro factor-shock)
- MissionCtrl — LLM hallucination detection via GRPO/RLHF-adjacent reward modeling · 8B model outperforms 70B zero-shot at 135× smaller param count
- Mergers-Acquisitions-Deal-Rater — Explainable ML deal-quality scorer · XGBoost + SHAP attribution trained on 70+ historical M&A deals (2018–2024) from SEC EDGAR filings
- Malware-Threat-Scoring-Engine — 0–100 composite scorer · SHAP interpretability · 100% accuracy on 200K held-out samples
- 🏆 Anvil 2026 × Scaler — Finalist / 3,200
- 🏆 Meta AI × HuggingFace OpenEnv — Finalist / 32,000+
- 📄 ICCI 2026 Published
Java · Python · Rust · TypeScript · SQL · Spring Boot · Kafka · Docker · PyTorch · XGBoost · ONNX · GRPO/RL · QLoRA · ClickHouse