Jump Trading

Going deeper on complex research questions is central to how we use AI at Jump Trading. Lucas B. recently spoke with AI Street about where AI agents are headed in investing, how to separate signal from noise, and why compute, data and infrastructure are prerequisites for research at the highest level. Read the full conversation below 👇

For 25 years, Jump has invested deeply in the technology behind our research. That's meant building our own hardware, developing large-scale research and simulation environments, working across petabytes of data, and continually experimenting with new approaches to computing. AI is central to that story. We’ve been exploring what Astra can do across our research environment with the kinds of problems we work on every day: broad questions without obvious answers, large and varied data sources, longer-running research, and work that can evolve over days rather than individual prompts. The ability to keep exploring matters to us; markets are complex and constantly changing, and some of the most useful research starts without knowing exactly where it will lead. Astra gives our researchers another way to go deeper. Hear more from Lucas B. below about what we’re learning, and explore more of our work across AI, machine learning and computing: https://lnkd.in/g9GPGiCE

Huge congrats to Mitesh Agrawal, Thomas Sohmers and entire Positron AI team on this recent raise! Jump Trading is a happy customer and early investor. While the rest of the industry scrambles for scarce HBM and advanced packaging, Positron AI's purpose-built inference ASICs go after the constraint that matters more and more — memory and power — squeezing far more tokens per watt out of every rack than general-purpose GPUs. As AI tips from training into high-volume inference, energy-efficient custom silicon stops being a nice-to-have and becomes the whole ballgame — and Positron AI is already shipping it into production. https://lnkd.in/g_DmnaCX

  Curiosity is the thread that binds us at Jump, and it might bring you here, too. That's the idea behind the Jump Trading Probability Cup, a world soccer forecasting competition hosted on SportsPredict.com this summer.    If you enjoy forecasting, probability, and difficult problems, this is your chance to put your skills to the test and compete against a community of sharp minds.    At the end, the top-performing eligible participant (18+) will earn a paid sports-related prediction markets fellowship at Jump Trading's Chicago headquarters. Registration is now open! Learn more about the probability cup and register here:  https://lnkd.in/g2wG7Aze

We meant to select six fellows this year. We selected eight. This year’s applicants didn’t just impress our panelists. In several cases, they forced them to rethink what’s possible at this stage of a research career. We’re proud to support the 2026–2027 Jump Trading Fellows. 👉 Learn more: https://lnkd.in/gV_TW4cz

Today, Jump Trading became one of the first financial services firms to deploy NVIDIA’s Vera Rubin NVL72 system — a rack-scale AI supercomputer built for frontier AI research. We’ve worked with NVIDIA for more than 15 years. In that time, the complexity of our research has grown dramatically. Markets are dynamic, adversarial, and constantly evolving. We run thousands of models across conventional and novel architectures, retraining continuously. Performance is determined by how fast we can experiment, test, and iterate. Our work goes beyond mainstream LLM inference. NVIDIA CUDA has given us the flexibility and programmability to explore ideas that would be difficult to replicate elsewhere. Vera Rubin NVL72 is the next step: greater compute density, stronger memory bandwidth, and significantly improved performance per watt. Which means more experiments, faster, at scale. We don’t treat infrastructure as a series of upgrades. We treat it as a long-term strategic platform. Each generation compounds. Jump is a deep learning research company in the business of predicting the future. We’re building an AI trading factory designed for curiosity, rigor, and hard problems across HPC, mathematics, and markets. We invest in what’s next. Then we build it. 👉 Learn more about our AI/ML work: https://lnkd.in/g9GPGiCE 👉 Read the full announcement: https://lnkd.in/g-E64z87

The next wave of AI breakthroughs won't come from scaling alone. They'll come from understanding intelligence itself.   We're partnering with UCL's Gatsby Computational Neuroscience Unit to fund 4 PhD scholarships starting in 2026.   The Gatsby Unit has spent decades asking foundational questions about learning, representation, and intelligent systems. The kind of questions that don't always have clean answers. The questions that tend to matter.                We're interested in how complex systems learn under uncertainty. Markets are one example. Brains are another. Both process incomplete information. Both adapt to changing environments. Both reward models that can separate signal from noise.   Our research and infrastructure are built around that reality.   This partnership builds on the Jump Trading Fellowships we established in 2025 and our broader support for research across mathematics, physics, AI, computer science, and the sciences.   👉 Discover how we're building the future of AI/ML at Jump, and who's building it with us: https://lnkd.in/g9GPGiCE   👉 Read about our partnership here: https://lnkd.in/ekcHZaNh