Quantitative researcher and research engineer focused on derivatives, volatility, fixed income, and statistical arbitrage.
I build and audit empirical research systems. My work emphasizes point-in-time data, executable timing, realistic costs, leakage controls, multiple-testing discipline, and reproducible metrics. When a result does not survive review, I publish the failure and correct the claim.
I am the founder and quantitative researcher at Navnoor Bawa Research LLC, a freelance markets analyst at FX Empire, and a Computer Science and Engineering undergraduate at Thapar Institute of Engineering & Technology (May 2027). Previously, I worked at Quant Insider and as a WorldQuant BRAIN Research Consultant.
Research writing | LinkedIn | Medium | YouTube | Email
| Program | Validation record | Current conclusion |
|---|---|---|
| Russell 3000 Statistical Arbitrage (site) | One continuous 11-window walk-forward with next-day (t+1) fills and per-stock costs. Newey-West, block bootstrap, Deflated Sharpe and Benjamini-Hochberg tests. Survivorship bias is disclosed. | No edge out of sample: pooled OOS Sharpe −0.33 (Newey-West p = 0.56); 1 of 35,491 pairs survives Benjamini-Hochberg; no better than the Gatev distance method (p = 0.88). The null result is retained. |
| Joint SPX/VIX Calibration and Volatility System (live) | Seven correction rounds (C16–C22): look-ahead, a circular feature, label noise, strike rolling, and a ~270-finding full-system audit. CI-tested, with mutation-checked look-ahead tests. | Heston fits the SPX surface to 0.9–1.4 vol points in-sample and 1.3–3.0 out of sample across five 2026 sessions, with vol-of-vol on its 2.0 ceiling in two. All P&L is mark-to-model; the VIX-options leg stays off until enough quote sessions exist. |
| WTI Crude Oil Volatility Research (live) | 926 purged OOS weeks (May 2008 – Sep 2026). HAR-IV compared with mean-reversion and persistence baselines; paired Diebold-Mariano test. | Volatility direction: 72.7% vs 66.6% for mean reversion (Diebold-Mariano z ≈ 4.1, HAC; ahead in 16 of 19 years). Level R-squared: 0.452 vs 0.173 for persistence. The leaked direction strategy was retracted. |
| U.S. Treasury Rates Monitor (live) | Direct Treasury XML, Federal Reserve H.15 history, and deterministic release checks. | Official daily CMT analytics and delayed CBOT futures proxies remain separate by design. |
Additional derivatives work: options pricing and strategy analyzer (BSM, Leisen-Reimer/CRR trees, Monte Carlo, SSVI/SVI local vol, American IV; 3,300+ tests in CI)
Winner, Trilemma Beta Global Data Science Tournament (2024) - probabilistic Bitcoin-return modeling in a field of 214 teams.
I treat the audit trail as part of the result. Every public claim should state its information set, execution timestamp, sample construction, missing-data treatment, costs, baselines, validation protocol, and known failure modes. I prefer a reproducible negative result to a profitable backtest that depends on hidden assumptions.
- Research: time-series analysis, statistical inference, walk-forward validation, factor models, volatility modeling, derivatives pricing, Monte Carlo simulation
- Engineering: Python, C++, SQL, NumPy, pandas, SciPy, PyTorch, scikit-learn, XGBoost, TypeScript, React, Linux, Git, CI
Open to quantitative research, derivatives, fixed-income analytics, and research engineering opportunities.