At AlgoQuant, I built equity mid-return models using NSE tick-by-tick data, with directional accuracy reported across 10-, 30-, and 60-second horizons. I also aggregated constituent forecasts for NIFTY spot and evaluated them in walk-forward tests.
I developed order-flow, passive-book persistence, and repricing features for return prediction, and built C++ and Python pipelines to reconstruct order lifecycles and extract research features. I also developed millisecond trade-co-occurrence labels and constructed cross-sectional long/short portfolios from lagged category-level order imbalances.
I built a Python and JavaScript market-data visualization platform with custom Canvas rendering, configurable subplots, formula transformations, and saved views. It also included parallel CSV loading, caching, and adaptive downsampling.
At Texas Instruments, I simulated receiver channels in MATLAB and compared models for BER prediction. My projects include a rainfall forecasting and alert pipeline, image restoration models, and a multi-memristor neural-network training simulator; I co-authored work presented at IEEE NVMTS 2024.
