How Jump Trading is scaling quant research with ChatGPT

Quantitative trading firm Jump Trading is utilizing OpenAI's GPT-6 Astra to scale its quantitative research and build predictive models for asset prices. According to Lucas Baker, Head of LLM R&D at Jump, predicting asset prices even slightly better than a coin flip at scale can create a successful strategy. Baker leads the development of agents and infrastructure that enable researchers to explore ideas with greater breadth and depth. Over the past year, AI at the firm has evolved from a tool for writing one-off code snippets into a versatile system capable of developing entire codebases and running advanced quantitative studies. GPT-6 Astra allows agents to handle complex workflows ranging from daily coding to hypothesis validation. Baker and his team treat AI like a colleague, where researchers define problems, environments, and evaluation criteria while steering agents in real time. With GPT-6 Astra, agents can merge and stack improvements recursively during long-running tasks that span multiple days and data sources, reducing the need for constant human intervention at every step. Because finance is a heavily regulated industry where mistakes carry financial and compliance consequences, Jump Trading relies on strong system design, boundaries, clear constraints, and human review to ensure workflows are safe. Agentic intelligence is also used to improve quality, security, and monitoring. Baker envisions a future where autoresearch and recursive improvement become standard parts of a quantitative researcher's workflow, blending human-defined priorities with autonomous agent fleets.