Claude Code ran real GPU experiments for Ångstrom AI
Ångstrom AI worked with the University of Cambridge and AstraZeneca to build CSP-MACE-Å, a model for predicting crystal structures in drug research. The model is presented as matching the accuracy of DFT, a slow and costly physics-based calculation method, while running about 10,000 times faster. Crystal structure prediction helps drugmakers find the different solid forms a molecule can take, because a later change in form can affect how a medicine works.
Ångstrom used Claude Code together with the anycloud during the research process. Claude Code launched groups of experiments, checked job status, downloaded results, and produced charts and summaries so researchers could decide what to test next. The work involved about 100,000 GPU jobs, mostly using cheaper across multiple s.
anycloud added controls so each Claude Code session could have limits on running jobs and total spend. When a limit was reached, new jobs waited while current jobs kept running, and Slack alerts showed cost, job counts, interruption rates, and blocked work.
Key points
- Ångstrom AI used Claude Code and anycloud to automate part of its research experiment loop.
- Claude Code launched GPU jobs, monitored them, pulled results, and created charts and summaries.
- The project ran about 100,000 GPU jobs, mostly on cheaper across multiple clouds.
- anycloud let the team set per- for live job spending and total budget.
- Slack alerts showed daily spend, job counts, interruption rate, and jobs waiting because a cap was hit.