Quantitative research on energy futures — mostly the WTI crude curve: calendar spreads, butterflies, relative value, and the market microstructure underneath them.
Six months on a systematic energy futures desk. Currently reading for two degrees: B.E. in AI & Data Science at RTMNU and a B.Sc. in Data Science at IIT Madras.
spreads-lab · live ↗ Built to teach myself how the WTI forward curve behaves. 421 exchange-quoted calendar spreads over 8 years of 1-minute data: a time-slider through the curve including the April-2020 super-contango, term-structure PCA into level, slope and curvature, a box/butterfly/cash-and-carry playbook, and a playable market-making game with Glosten–Milgrom adverse selection. A study, not a strategy.
mm-lab Market-making simulator. Quote around a jump-diffusion fair value; uninformed flow pays you the half-spread, informed flow arrives ahead of the jumps. P&L decomposes into spread capture, inventory risk, adverse selection and liquidation — an identity asserted to 1e-6 across 34 tests. The finding worth the build: the naive quoter captures more gross spread than Avellaneda–Stoikov and keeps none of it.
crude-butterfly-spreads Streamlit backtester trading six butterflies simultaneously along the curve, with thresholds adjustable live so the sensitivity is visible instead of hidden.
pca-statistical-arbitrage Rolling PCA strips the common factor from crude spreads; the residual trades on multi-horizon z-scores, with an Ornstein–Uhlenbeck half-life turning the signal into a holding period.
market-regime-detection Shannon and Rényi entropy, Lyapunov exponents, correlation dimension and Markov transitions — used to size and gate other strategies rather than to trade on their own.
wti-opening-range-breakout Session-aware ORB with VWAP-relative bias classification and MFE/MAE trade-path analysis. Results are withheld pending a fill-logic fix — the README says exactly why.
Python, C#/.NET, SQL. Market-data pipelines, backtesting engines that don't lie to you, and execution and pre-trade risk plumbing.