Over the past four years Orbital Industries has been developing new advanced hardware with AI — everything from discovering new advanced materials, engineering new hardware and optimizing manufacturing. CurieOS is the culmination of that hard-won experience: it's our AI operating system used across the entire company to discover materials, engineer hardware, and optimize the running of our labs.
Bringing AI into an industrial company surfaces three big problems. The first is that, while AI has the equivalent of a PhD in every scientific discipline, it only delivers when it's given full organizational R&D context. The second is that scientific workflows are unusually complex and long-running — requiring new ways to orchestrate and manage fleets of agents. And the third is that scientists and engineers need a new and intuitive work surface to collaborate with AI: instead of disparate reports or spreadsheets, they should collaborate with AI using on-demand scientific software connected to their lab and instruments.
Each one runs continuously in its own secure, GPU-accelerated environment with the full scientific computing stack, including Orb, our atomistic simulation model. Agents work as co-pilots alongside our scientists, or in autonomous swarms orchestrated by the OS, taking on R&D workloads at scale.
Beneath the agents, CurieOS holds the whole organization — experiments, test rigs, requirements, materials, suppliers — in one live, connected picture: a context graph. Everything an agent does is checked, permissioned, and recorded there, so AI conversations stop being throwaway and start compounding into organizational capability:
Non-technical users stand up sophisticated, secure, team-wide scientific tools in natural language, within hours: simulation campaign dashboards, bespoke hardware digital twins, live experiment trackers, procurement dashboards, each with deep AI integration. Every new tool adds context to the graph instead of becoming a silo, connecting work between teams and enabling organization-level reasoning.
CurieOS scales an individual analysis into a project-scale campaign: high-throughput computational screening, experimental optimization loops, multi-stage computational chemistry pipelines, running for days and pausing for human review where it matters. When a new experiment lands, it runs a screening campaign overnight to suggest the next best experiment. When a test rig produces its data dump, an agent extracts, cleans, analyzes, and loads it. Digital twins retrain every night on the latest data.
Because everything shares one graph, a change anywhere propagates everywhere. Change an engineering requirement and CurieOS re-runs the computational studies, updates designs and digital twins, and adjusts procurement, collapsing systems-engineering cycles from months to days.
CurieOS connects to the systems where an organization's knowledge already lives, and learns from them, improving the predictions of the models that run on it. It's how we're developing two-phase cooling at Orbital Industries. The working fluid, the cold plates, and the manufacturing process that builds it are three separate disciplines sharing one context graph.
When new hardware specifications update the operating conditions of the cooling loop, CurieOS agents re-run the cold plate thermal and flow simulations. Similarly, when the cold plate geometry changes, manufacturing constraints can move with it.
If you are interested in discussing your project and collaborating with Orbital Industries, please contact us below.
Contact us
