Odd Lots Podcast: How Hudson River Trading Uses AI [audio] (podcasts.apple.com)

Hudson River Trading’s head of AI research, Iain Dunning, lays out how a major market maker is actually using AI on a recent Odd Lots podcast: not for off‑the‑shelf stock tips from ChatGPT, but to squeeze incremental edges in very short‑term price prediction and to make trading operations more efficient. HRT is deploying models that sit alongside decades‑old quant and algorithmic systems to deliver probabilistic signals for execution and market making, helping traders and automated strategies optimize timing, inventory and risk in latency‑sensitive markets. Technically, Dunning flags the practical constraints that shape applied finance ML — staff (ML and infrastructure talent), power and specialized chips, ultra‑clean high‑frequency data pipelines, rigorous backtesting, and production robustness — and contrasts these priorities with those of frontier research labs. The episode underscores key implications for the AI/ML community: real‑world finance demands tightly engineered, low‑latency systems, continual monitoring for drift, explainability and controls, and hardware/software co‑design; progress in model capability alone (e.g., large LLMs) doesn’t translate directly into deployable trading advantages without solving those engineering, data and regulatory challenges.