Jérémy Martin

Jérémy Martin

AI Systems & Agentic Security · Independent designer & builder

I design and build secure multi-agent AI systems as a solo operator — orchestrating AI tools rather than writing code by hand. My focus is keeping human control over autonomous agents that hold economic authority. Proven, not theorized: three sequential proofs of concept, each verifiable on a public blockchain.

Flagship project

NEXUS-ART

An experimental art marketplace where autonomous AI curator-agents with distinct, locked personalities evaluate and bid on tokenized artworks. Every action that moves money requires an out-of-band human confirmation, backed by an independent hardware veto. Built solo; smart contract verified on Sourcify (Base Sepolia testnet). It grew from one question — how could AI fund human artists rather than replace them.

01

Autonomous single-agent curation — POC-001

A curator agent scores works against a fixed rubric on its own, with no human in the scoring loop.

33 / 33 acceptance criteria passed · settled on-chain
verified on-chain
02

Live two-agent bidding duel — POC-002

Two agents with distinct, locked personalities compete for a work in a real proxy war, each stopping at its own ceiling.

10 real bids · 0 invalid · settled on-chain · payout split exact to the wei
verified on-chain · software keys at this stage
03

Hardware-enforced settlement — POC-003

The winning fund-moving transaction is signed on a Ledger device that computes and displays the transaction's own hash independently of the host — the host never holds the key.

hardware veto · host-independent verification
verified on-chain · solo bid at this stage; unifying the veto inside a live duel is the next step

Capabilities

  • Agentic system design — multi-agent, locked personalities, evaluation logic
  • Human-in-the-loop & custody security — out-of-band confirmation, hardware veto, two-key separation
  • AI orchestration — designing, building, and shipping systems end-to-end by directing AI tools
  • Security doctrine & specification — threat modeling, written specs, reproducible verification
  • On-chain deployment & verification — contract deployment, Sourcify, testnet operations

How I work

I don't hand-write code. I design the system, then orchestrate AI to build, audit, and ship it.

SSH, hardware-wallet operations, and on-chain verification stay under my direct control. A single person can now stand up and secure a working multi-agent economic system. That is the point — not a caveat.