Building the infrastructure that AI actually runs on.
AI and distributed systems engineer, working on the layer most people never see: an operating system for machines and data centers, inference routing across heterogeneous compute, confidential computing on real hardware, and settlement infrastructure for assets worth hundreds of millions. Architecture through to production.
01
Selected experience(5)
The work that is live now. The full chronology, including music and public service, is on the about page.
01
Living Worlds Lab
An AI lab building persistent worlds where AI workers live, work, and change only on evidence.
- Building the Living Worlds runtime: a persistent universe in which selves and populations of selves live continuously across worlds.
- Designed a world model rather than a memory: state with a current value, history and rollback, and expectations settled by what actually happens.
- Built the gate at the core of the runtime: a self's beliefs move on ratings and outcomes, but its practice changes only when an independent check beats the old way.
02
Praecise
An AI and infrastructure company. We own the machines, the operating system that runs them, and the cloud placed on top.
- Created Furcate OS, a Linux-based operating system designed to run machines and data centers.
- Runs the machines end to end: measurement, forecasting, placement, power and thermal control across the hardware Praecise owns.
- Building the cloud layer that places and supervises workloads on machines we own and operate.
- Praecise operates as a cloud service provider, delivering digital infrastructure powered by data centers and machines running Furcate OS.
03
Tenzro
A decentralized AI and distributed computing network, supplying AI compute to data-center operators worldwide.
- Founded and architected a decentralized compute network combining distributed execution, secure identity, multi-VM execution and settlement.
- Built execution frameworks for workloads that coordinate, transact, and settle across the network.
- Implemented privacy-preserving computation using Trusted Execution Environments and secure infrastructure.
- Developed distributed AI inference routing to optimize cost, latency, and performance across heterogeneous compute.
- Built native settlement for the compute and inference the network runs, with cross-chain interoperability.
04
PRVNZ
The trust layer for creative work: licences for music, film, visuals and writing, checked on the machine where the work happens.
- Issues licences for the creative and entertainment industry and checks them on the machine at the moment of opening, with or without a connection.
- Built passports that travel with an asset and record who made it, what it is, and that nothing has changed since.
- Fingerprints recordings from the sound itself, so an owner's word on AI training can be looked up in any file.
- Machine Sounds (machinesounds.xyz): guitar and amp sound processing: amps, pedals and effects, tone captures made with machine learning, and the systems that run models on guitar signals.
05
Boli
Infrastructure for real-world asset tokenization and payments in stablecoins, crypto and fiat. Founded as a Swiss non-profit.
- Providing tokenization infrastructure for more than $500M in assets across resort and real-estate portfolios.
- Designed systems that abstract asset digitization, on-chain integration, and cross-system settlement.
- Technology partner for the Government of Maldives’ smart city project.
- Operates payment infrastructure across stablecoins, crypto and fiat, including cross-border payments in partnership with Bridge and Stripe.
- Worked with governments and institutions on programmable asset infrastructure.
02
Writing
Essays on intelligence, infrastructure, and what runs underneath.
01
Permissionless Intelligence
Most countries can't build a gigawatt AI campus, and most operators can't rent one. Open models are now good enough that they don't have to — what's missing is a network anyone can join, run a node on, and provide AI to. Notes on building Tenzro, and why it matters for small nations and independent operators.
02
The Gating of Intelligence
When access to the most capable AI models becomes conditional, gated, and revocable overnight, the case for verifiable and distributed systems stops being ideological and becomes practical.
03
Who Does the Agent Answer To?
AI agents can now hold money and spend it — the payment rails arrived this spring. Identity, liability, and governance are being retrofitted afterward, and that order is backwards.
04
Privacy Is Not a Setting
A court order for twenty million chat logs, an opt-out nobody saw, and a cloud vendor that couldn't promise anything under oath. Data sovereignty isn't a toggle you flip — it's decided by architecture and jurisdiction, long before you reach the settings page.
03
Core expertise
Where the work concentrates.