Shahram Nikbakhtian

Designed, built, and operated a fully automated HFT crypto market-making and options system across Binance, Bybit, BitMEX, Coinbase, and FTX. $1bn+ cumulative volume; sub-millisecond latency capable of 1000 trades/sec. End-to-end low-latency C/C++ with no external libraries — exchange connectivity, networking, pricing, inventory risk. Lightweight C websocket server brokering between the algo processes and a JS web dashboard / Python Telegram control app. Two independent monitoring servers with SMS / WhatsApp / Telegram alerts. Mobile git-push to live in ~10 minutes.

Research and production implementation of hybrid interest-rate / FX models combining local and stochastic volatility. First in the finance industry to port hybrid pricing to CUDA GPU — orders-of-magnitude speed-up for exotic options risk.

Led the AI/ML work behind a peer-reviewed UK Biobank study showing that bedtime timing is associated with cardiovascular risk. Built deep-learning models extracting latent sleep states from terabyte-scale raw accelerometer signals. Covered by BBC, The Washington Post, and The Guardian.

Press: BBC · The Washington Post · The Guardian

Built two complementary smartphone-camera heart-rate techniques: (1) facial colour change with Fourier analysis of the red/green channels, and (2) Newtonian head-motion induced by blood influx after each heartbeat, using front and rear cameras simultaneously to cancel hand-shake. Correlation-based signal periodicity to separate the true cardiac component from motion noise — accurate enough for daily monitoring.

Deep convolutional networks on smartphone motion sensors predicting falls in elderly patients. Separately, YOLOv3-based wound-area measurement pipeline for venous leg ulcer management — first deployment of YOLOv3 to iOS / Android (2018), and unique among competitors in measuring wound surface area without requiring a physical reference object in frame.