My career in four acts — sensors, chains, community, agents. Everything below actually shipped.
Read the record ↓Before IoT was a slide-deck word, my job was putting a network inside rubber — and making a fleet trust what it says.
On it from 2015 — the tire becomes an input device: in-tread sensors feeding ABS and traction control. World-first 5G transmission, then standard fitment on the McLaren Artura.

Fleet tyres as a modular IoT platform — in-tire sensors over Bluetooth, Pirelli Cloud forecasting fuel and maintenance, integrated where fleets already live: Infogest GoliaWeb, Geotab Marketplace, Autotrac in Brazil.

Contribution to the winning bid for the EU's innovation community for city mobility — vehicles as roaming sensors for the roads they drive.
Tyre-as-a-service: tyres and roadside assistance by monthly subscription, from an app. Direct-to-consumer for a dealer-first company.

The invisible layer: ML wear estimation without measuring the tread, and sensor auto-location without manual pairing. The models were the product.
Crypto's infrastructure years: I turned blockchain capability into things developers could actually ship on.
No-code tokenization — drag-and-drop tokenomics, generator, explorer, wallet. The bet: the next million token creators won't be Solidity developers.

Horizen's EVM sidechain: the full Ethereum toolchain on ZK-secured infrastructure. Hardened on testnet with public bounties, then mainnet.
The first blockchain purpose-built for verifying ZK proofs — attestations usable across chains at ~90% below L1 cost. A new category: verification as its own layer. My part started in 2021 — business case and strategy.

Hand the power over: SDKs to build with, audits anyone can verify, votes that count, hackathons that pull new builders in.
Sidechains at ~1,000 TPS as something you instantiate, not something you found a company for. Shipped with Zendoo's mainnet activation.

The official TypeScript SDK for VeChainThor — 100,000+ weekly downloads on npm (sdk-core alone). The blockchain's power, importable in one line.
The first privacy-preserving audit blockchain, with Celsius: zero-knowledge proof-of-reserves in near-real-time — before the industry learned why it mattered.

Governance as a product surface: validators and Stargate NFT holders vote gasless via relayers; a free no-code Proof-of-Participation module onboards non-crypto users at zero cost.

The developer front door: developer tools and templates at vechain.org/builder-hub, and the $30K DoraHacks global hackathon that ran through October 2025, pulling new builders and projects into the ecosystem.

My job now: making AI agents into things people can build, evaluate, trust — and buy.
Build, test and publish AI agents: Studio for creation, Evaluate for quality and security gates, Canvas for UI inside the chat, self-improvement in the creator's language. Shipped backend to pixel, in 8 languages.


Three AI colleagues in the org's Slack, built as n8n workflows: coordination, content intelligence, content ops. Prompts versioned as markdown, deployed via MongoDB, one shared self-improving memory.


Twenty repos of my product experiments — the live ones are underlined.
MediBook: every specialist gets a WhatsApp AI agent trained on their own documents — booking, calendars, invoicing, payouts in one product.






























Bet on the capability — pick the new primitive early, let it decide the product.
Evaluation is UX — people adopt what they can verify on screen.
Speak the user's language — localization is respect, not overhead.
Demo mode is a feature — the complete flow is the pitch.
Shipping is a system — process built so launches land calm, and keep landing.