When Anish Mariathasan watched robotics teams at Harvard's machine-learning lab wrestle with design changes, he noticed the same pattern repeating: a mechanical tweak would ripple through circuit boards and firmware in ways no one anticipated until integration testing, often weeks later. The fix-it costs piled up fast.
Now Mariathasan and co-founder Ayush Patel think they've built a way out of that mess. Their startup, Osseus, emerged from Y Combinator in summer 2026 with a platform that stitches together CAD files, circuit schematics, GitHub repositories, and project-management boards into what they call a "live dependency graph." The San Francisco company uses AI agents to trace how a single design decision cascades across an entire robot build and to suggest fixes before hardware ships.
The problem Osseus is tackling isn't new, but it has grown more acute as robotics startups proliferate. Mechanical engineers live in CAD software, electrical engineers inhabit schematic tools, and software teams work in GitHub. Information flows between these worlds haphazardly, usually through Slack messages or weekly standups. According to a claim made on the Y Combinator launch page for Osseus, catching a requirements error during final integration can cost up to 78 times more than if spotted early in the design phase.
Osseus plugs into existing tools rather than asking teams to abandon their workflows. Once connected, the platform monitors the dependency web. Swap a motor specification, shift a mounting bracket by a millimeter and a half, or push a firmware update, and Osseus flags which requirements, tests, and downstream systems might break. The AI agents then run simulations, draft proposed fixes across both hardware and software repositories, and validate the changes before an engineer ever sees them.
In one demo trace the company shared, moving hexapod leg centers from 108.5 mm to 110 mm outward triggered automatic updates in two C++ header files, complete with staged code patches and validation tests tied back to the source CAD geometry. Engineers review and approve before anything merges, but the grunt work of tracing dependencies happens in the background.
The platform also centralizes design-review conversations. Threads from mechanical, electrical, and firmware teams appear in a single interface, and supplier comments remain searchable. Visual diffs on 3D assemblies show exactly what changed between revisions, with annotations tracking each modification.
Osseus has woven a components marketplace directly into the design workflow, an unexpected feature for a traceability tool. The marketplace pulls real-time part availability and lead times, then evaluates designs against what teams can actually source. When a specified component faces extended delivery windows, the system suggests compatible alternatives. Marketing screenshots show example listings like "CubeMars AK80-9 — $720 (buy)" and "RMD-X8 Pro — $34/day (rent)" nested within a motor-selection widget, though Osseus hasn't disclosed which distributors or inventory sources power the marketplace.
Mariathasan co-founded a compliance AI startup before Osseus and conducted machine-learning research at a joint Harvard-Hong Kong University lab. He holds a biomedical engineering degree from Imperial College London. Patel interned twice at Apple and built an agentic quality-assurance system; he has a computer science degree from Imperial as well. The Y Combinator directory listed the team size as two at launch. Jared Friedman is the primary partner shepherding the company through the accelerator.
Osseus enters a landscape where competitors have carved out adjacent niches. AllSpice offers ECAD collaboration with an AI design-review agent called DRCY, described in a peer-reviewed paper published earlier this year. Valispace links requirements to engineering values and updates them automatically when designs shift. PTC's Windchill and Codebeamer provide OSLC-based traceability connecting requirements, CAD, and Jira for enterprise hardware teams. Foxglove builds visualization and data platforms aimed at robotics development. Solus Forge, another robotics workspace tool, remains in waitlist mode.
Data from Dealroom lists Osseus as having raised $125,000 from Y Combinator, though the company hasn't confirmed funding details publicly.
On data governance, Osseus has emphasized that customer data won't train foundation models and that on-premises or private-cloud deployment options are available. The company says it will provide documentation for SOC 2, ISO/IEC 27001:2022, ISO/IEC 42001:2023, HIPAA, and GDPR compliance on request. Its trust center showed all systems operational at the time of launch.
Pricing remains undisclosed, but Osseus outlined three tiers on its website: Builder for individuals, Team with shared workspaces and permissions, and Enterprise with single sign-on and custom integrations.
Whether the platform can convince robotics teams to adopt yet another tool in an already crowded stack will depend on how well it delivers on the promise of catching expensive mistakes early. For now, Osseus is betting that the pain of fragmented toolchains has become acute enough that teams will welcome a unified view, even if it means adding one more login to the mix.
