Independent data scientist and quantitative researcher in Suwanee, GA. I've spent 10+ years building predictive models and forecasting systems across macro, options, time series, and prediction markets. Active research right now is in reinforcement learning and frontier reasoning benchmarks like ARC-AGI.
MacroForecastbyCM is a live, daily-updated dashboard with house forecasts and Kalshi market consensus for every major US macro release: Core CPI, Core PCE, NFP, ADP, Unemployment Rate, and Weekly Jobless Claims. It also keeps an honest track record of how my forecasts have done against Kalshi and the actual prints. Source at github.com/ChiragMirani/macro-dashboard.
Gita is a small web app for reading the complete Bhagavad Gita verse by verse in English. All 700 verses, public-domain translation, free to read and share, installable on iOS and Android home screens. Source at github.com/ChiragMirani/gita-quotes.
Shree Hanuman Chalisa (English) is a free iPhone and iPad app for reading the Hanuman Chalisa verse by verse in plain English, with the Hindi and the Hindi in English letters one switch away. No ads, no account, nothing collected, and it works offline. The same text is also on the web at chiragmirani.github.io/hanuman-chalisa.
On Kaggle as cmirani for 9 years. Silver-medal notebook on a leakage-free time-series training and test pipeline (77 upvotes). Recently finished top 22% (379 of 1711 teams) in the MITSUI Commodity Prediction Challenge. Currently competing in ARC Prize 2026 (ARC-AGI-2) and the NVIDIA Nemotron Model Reasoning Challenge.
A focused subset of in-production research. Code, methodology, and live PnL stay private. Happy to discuss under NDA.
-
US macro release-day trading stack. House forecasts, surprise classifiers, and prediction-market consensus calibration for Core CPI, Core PCE, NFP, ADP, UR, and weekly jobless claims. The public proof point is the live dashboard and its honest track record. The bet sheet and position layer stay private.
-
Term-structure butterfly trading across UST, SOFR, and EUR rates. Z-score and carry-driven 2s5s10s, 2s3s5s, and broader fly engines built on Bloomberg, with multi-curve cross-sectional scoring and regime overlays.
-
Prediction-market alpha on Kalshi and Polymarket. Real-time monitoring on 100+ series, with delta engines, cross-platform price discovery, and event-calendar-aware alerting around Fed decisions, macro prints, and market-moving headlines.
-
0DTE SPX program. Live short-dated option strategies (iron flies, delta-hedged structures, calendar and ratio overlays) with automated MES futures hedging through IBKR. Vol-premium-aware sizing with intraday risk controls.
-
Cross-asset factor stack across TLT, QQQ, GLD, and USO. Multi-asset signal generator combining macro, seasonality, and trend or persistence factors, backed by an expanding-window backtest grid that has evaluated 170K+ weight configurations.
Exposure-Time Normalized Performance: Strategy Evaluation Under Binding Exposure Constraints (Feb 2026). Formalizes exposure-time as a measurable input in strategy evaluation, defines exposure-normalized Sharpe and MAR ratios, and shows in Monte Carlo that exposure-normalized rankings reverse versus raw Sharpe in 69.5% of simulations. SSRN.
A Predictive Model for 10-Year Treasury Futures (independent research, 2025). PDF.
Open to consultation, collaboration, or trading-research conversations. Find me on LinkedIn, X, or Kaggle.