● BUILDING
TRACER
Modernise legacy circuits
A verifiable copilot that reverse-engineers legacy circuits step by step. The wedge. Core primitive proven at AMD; MVP in development.
● BUILDING SUBSTRATE · PROBLEM VALIDATED
AIagentsforthephysicalworld.
The agentic platform for deep-tech engineering.
↓ scroll
01 / THE GAP
SOFTWARE
Cursor. Copilot. Devin. Lovable.
Multi-billion-dollar AI companies. Every fund has a thesis; every grad with an LLM API key is launching one.
HARDWARE
Nothing.
The engineers designing chips, satellites, defence systems, energy grids, and medical devices still work in tools from the 1990s. The AI revolution skipped them.
To build the platform for them, you need a founder fluent in both worlds. The intersection is empty.
02 / SUBSTRATE
The agentic platform for deep-tech engineering.
Tracer first. The other agents are the direction — each unlocked by the data and engineer trust the previous one earns.
DIRECTION
SCHEMATIC
Design new circuits
Generates new circuit designs from functional intent.
DIRECTION
EMBEDDED
Low-level firmware
Writes and verifies embedded code against real hardware.
DIRECTION
MECHANICAL
3D / CAD generation
Produces parametric mechanical designs from spec.
DIRECTION
SILICON
Chip design & verification
Agentic RTL design, synthesis, and verification.
DIRECTION
ASSEMBLY
Chip packaging & assembly
Optimises packaging, layout, and assembly workflows.
What Cursor is to software, Substrate is being built to be for the rest.
03 / THE WEDGE
We start with Tracer.
A verifiable copilot that helps engineers modernise legacy electronic circuits — step by step, engineer-in-the-loop.
01
Existing budgets, today.
Defence reshoring and supply-chain mandates have already allocated procurement budgets for modernisation work.
02
No behaviour change required.
The engineers I interviewed told me this directly — they have the problem, and they want a tool that fits inside their existing EDA workflow.
03
AI-tractable today.
Multimodal frontier models can finally read schematics, datasheets, and CAD in the same pass. The AMD core primitive proves it works on real circuits.
04
The work compounds.
Each modernisation engagement adds proprietary design data and engineer trust — the foundation the next agent in the roadmap is unlocked by.
Tracer is the wedge. The platform compounds underneath as engagements accumulate.
04 / WHAT'S PROVEN
The hardest part is proven. The plan is grounded in real engineer pain.
2nd
OF 100+ TEAMS
AMD Pervasive AI Contest
Team of 4. I led and built the circuits parser and core circuit-reasoning logic — the primitive Tracer is being built on. Codebase and trained artefacts available.
150
USER DEPLOYMENT
Argus (prior venture)
Multi-agent AI I built solo and deployed across 180DC Southampton. Proof I can take agentic systems from prototype to real users.
£3K
FUTURE WORLDS PRIZE
Argus, pitched and funded
Argus won the Future Worlds Enterprise Prize and was pitched at Future Worlds. Proof I can build venture-grade product and pitch it to investors.
1st
OF ~15 TEAMS
BAE Systems Hack the Future
End-to-end hardware + networking + backend + UI in 24 hours. The hardware fluency almost no AI founder has.
DISCOVERY
What engineers told me.
I interviewed practising engineers at firms specialising in modernisation, and at defence and biomedical companies that also do modernisation work.
The signal was consistent: legacy circuit and system modernisation is real, painful, and they urgently need a faster way to do it.
“Don't give us an autonomous black box. In our domain that just multiplies errors. Give us a tool that works step by step alongside the engineer, so we can verify every step.”
Paraphrased from the engineer interviews. It is the design brief for Tracer: a verifiable copilot, not autopilot.
05 / WHY NOW
AI finished software. Deep-tech is what's next.
01.
Multimodal frontier models can finally read schematics, datasheets, and CAD in the same pass.
The visual half of engineering is finally tractable.
02.
Agentic orchestration crossed the reliability line.
Tool-using, multi-step agents are now production-grade. The piece needed to make a verifiable copilot feel real.
03.
Defence reshoring and American Dynamism made this commercially urgent.
Real budgets are being allocated this fiscal year for legacy modernisation, secure supply chains, and domestic chip capacity.
The window is open. It won't be for long.
06 / WHY ME
Most AI founders can't read a schematic.
Most hardware engineers can't ship an agentic system.
I do both.
Second-year EE at Southampton · Zepler Prize · 2nd at AMD Pervasive AI (led 4-person team, built the core) · 1st at BAE Hack the Future · Argus built solo, deployed to 150 users · Tracer validated through engineer interviews