WhoMakesIt · Startwell Ventures · ai / supply chain analysis
What could we make in the UK?
An AI supply chain analysis platform connecting UK manufacturing capability to buying decisions, built with the winning Startwell team at NVIDIA Hack for Impact.
Product experience design & frontend engineering · Built with the Startwell team
Hackathon winner · Sheffield, November 2026
Data visualisation · AI enablement · Product & service design · Engineering & systems
A shared Startwell team achievement: Danny Tomalin designed and engineered the product experience; Ben Eastwood designed and built the AI system and backend; Gary Neville shaped the impact story. Together, we won NVIDIA Hack for Impact in Sheffield and took home a DGX Spark.
- NVIDIA Hack for Impact winner
- Supply chain analysis & product engineering
- Local AI on DGX Spark
The problem
The UK can make far more than it knows. Manufacturing capability is scattered across trade records, company data and manufacturers’ websites. Buyers and regional decision-makers need a way to connect that evidence to the parts they need, and understand where a UK supply chain could work.
My contribution
I designed and engineered the product experience. Ben Eastwood designed and built the AI system and backend that powered it. Gary Neville shaped the impact story and why it matters. We brought it together as the Startwell team.
The key decision
The tree connects a product to its parts, manufacturing steps and potential suppliers. Agent activity, analysis, location and the alternative bill of materials offer different views of the same investigation. That lets a buyer move from the whole supply chain into a specific part and the evidence behind a proposed match. The final decision remains with a person.
The result
Together, the Startwell team won NVIDIA Hack for Impact in Sheffield on 13 November 2026, taking home a DGX Spark. The win gives us hardware to keep developing and demonstrating the approach. This is a working prototype; it has not yet been validated on a real buyer’s parts list.
Explore the approach & technical detail
Making the agents’ work usable
The demo moves from a hydraulic cylinder to a reactor coolant pump for a small modular reactor. I designed and built the experience that connects the parts, agent activity, supporting evidence and manufacturing routes. Buyers can explore the supply chain, compare route options and inspect an alternative bill of materials; regional decision-makers can see capability gaps and import opportunities.
From a parts list to a supply chain
Give WhoMakesIt a bill of materials and its agents investigate who could make each part, challenge the evidence and check production steps. The product connects that analysis to supplier maps, routes planned with NVIDIA cuOpt, an alternative bill of materials and a supply chain report. The demo combines NVIDIA Nemotron, Nemotron Embed, cuOpt and NeMo Agent Toolkit. Outputs show where capability is indicated, conditional, unproven or where no UK source was found, so a potential match is not presented as a confirmed supplier.
Local compute, human decisions
The team built the system to run open-weight models on NVIDIA DGX Spark. Local inference supports keeping a buyer’s parts list within their own computing environment. Supplier discovery still reads public web sources, and potential matches need human review and supplier confirmation.
Why it matters
The demo sets out three connected ambitions: keeping jobs and manufacturing value in the UK, shortening and securing supply chains, and growing sovereign AI capability. It shows how procurement frameworks can affect the suppliers a buyer can choose, and turns the analysis into a downloadable alternative bill of materials and a supply chain report intended to inform government policy. These are proposed uses and ambitions, rather than measured economic or policy outcomes.
