AI and robotics system design
Feasibility and system design for agentic AI and robotics products, with safety and regulatory readiness built in from the start rather than bolted on.
Who this is for
- Founders building an AI agent or a robot who need to know what is feasible before they commit a team to it
- Engineering teams moving a prototype towards a product that has to work at a customer site
- Companies whose product will be regulated and want compliance designed in, not retrofitted
What we do
- Assess feasibility: what the technology can do today, what it costs to run, and where it will fail
- Design the system: agents, tools and human-in-the-loop for AI; sensing, control, fleet and operations for robotics
- Map the safety and regulatory path (CE marking, ISO standards, EU AI Act) into the architecture and the roadmap
- Plan deployment, operations and evaluation so the product survives contact with real users
- Prototype or review code where that settles a question faster than a document
What you get
- A feasibility report with a clear yes, no or 'yes if', and the evidence
- System design documents and diagrams your team can build from
- A compliance path with what you can self-assess and what needs a notified body
- An evaluation plan and the first metrics to track
How it runs
- Format
- Fixed-scope project with a workshop at the start and a handover at the end.
- Typical length
- Two to six weeks.
- What I need from you
- The use case, the constraints, existing prototypes or code, and the people who will own the system afterwards.
Where I've done this
I led the engineering and CE compliance of care robots at Bearcover (ISO 13482), deployed and operated autonomous sweepers for customers at Enway, built the robotic system and docking vision for VoltVogel's EV-charging robot, and led an AI team shipping a document-intelligence pipeline at Carbon Freed. My own research on humanoid locomotion with reinforcement learning and on agentic systems is on the Insights page.
Questions founders ask
- Single agent or multi-agent?
- Usually single, with good tools, until there is evidence you need more. Multi-agent systems are expensive to evaluate and easy to oversell; I will show you the data before recommending one.
- Can you cover EU AI Act readiness?
- Yes, as part of the design: risk classification, the documentation you will need, and what you can self-assess. I will be direct about where a notified body or a lawyer is needed.
- Do you do hardware?
- I design the system around the hardware and I have deployed robots at customer sites, so I know what breaks there. Mechanical and electrical design itself I bring in partners for.
- What if the feasibility answer is no?
- Then you get the no, the reasons and the nearest thing that would work. That report has saved more money than any other deliverable I produce.
Start a conversation
Tell me what you're building and what's in the way. I reply within one business day, usually with a first thought, not a sales pitch.