Trusted and reproducible workflows for machine learnt interatomic potentials
On 8th January 2026, the Physical Sciences Data Infrastructure (PSDI) hosted a webinar titled “Trusted and reproducible workflows for machine learnt interatomic potentials”, presented by Elliott Kasoar and Dr Alin-Marin Elena. This webinar explored machine learnt interatomic potentials (MLIP) software frameworks enabling scalable, accurate simulations, benchmarking, and workflow integration.
For those who missed the live session or wish to revisit the discussion, the full webinar is available on our YouTube channel.
Stay tuned for future PSDI webinars by subscribing to our YouTube playlist: PSDI Webinar Series.
Abstract: This webinar explored recent advances in machine-learned interatomic potentials, highlighting software frameworks that enable scalable, accurate atomistic simulations. It introduced janus-core, with emphasis on aiida-mlip for provenance-aware HPC workflows, and ML-PEG, an interactive benchmarking and usability guide for developing, testing, visualising, and integrating MLIPs within broader PSDI ecosystems and data.
