Research & Development
What we are building before clients ask for it.
A standing research programme in physical AI, adaptive manufacturing and industrial agents — funded internally, applied commercially, and published where we can.
Programmes
Six active threads.
Physical AI
Learned manipulation policies for contact-rich tasks — insertion, deformable material handling, adaptive grasping — trained in Isaac Lab with quantified sim-to-real transfer.
Adaptive manufacturing
Cells that re-plan from an updated CAD model rather than a manual re-teach. The objective is changeover engineering measured in days, not weeks.
Industrial AI agents
Agents that plan over plant state within bounded authority — proposing schedule changes and maintenance actions with every decision logged and reversible.
Twin drift detection
Measuring divergence between a commissioned digital twin and the physical cell, using it as an early indicator of mechanical degradation.
Humanoid & general-purpose robotics
Evaluating general-purpose platforms for high-mix material handling where fixed automation cannot justify its tooling cost.
Synthetic data for inspection
Generating labelled defect imagery for classes too rare to photograph, with measured transfer to real-part detection performance.
Method
Research with a commissioning date.
Every programme carries an application hypothesis and an exit criterion. If a thread cannot show a path to a production line within a defined horizon, it is closed and written up rather than quietly continued.
We publish methods, benchmarks and failure modes. Negative results are the most undersupplied thing in industrial AI, and the most useful to a plant engineer deciding where not to spend a budget.
Collaboration
How to work with the R&D group.
Sponsored research
A defined question, a fixed term and a written deliverable. You own the results in your domain; we retain the general method.
Joint development
Co-funded development of a capability neither party would build alone, with agreed IP allocation set before kickoff.
Academic partnership
Supervised projects with engineering institutions, focused on problems drawn from real production constraints.
Next step
Have a problem nobody has solved for you yet?
Contract R&D is how most of our long engagements begin: a bounded feasibility question, answered with data rather than a proposal.