Index

vaN


Infrastructure · Bittensor · AI

15 years in infrastructure. The last two spent building in the Bittensor ecosystem and AI agent space. Based in Auckland, working at the intersection of cloud systems and decentralised AI.


AUCKLAND, NZ · --:-- NZv2026.1AVAILABLE FOR WORK

4 PUBLIC REPOS · LAST SHIPPED yesterday

Projects

  • zhen-subnet
  • Python
  • ★ 0
  • 2026-04-15

zhen-subnet

A decentralised calibration layer for digital twins.

Zhen is a domain-agnostic platform where independent parties compete to calibrate simulation parameters against ground truth. The first vertical is building energy, with RC thermal models scored against the BOPTEST emulator using ASHRAE-standard metrics. Running on Bittensor testnet as subnet 456, with three milestones proven: two-model architecture, competitive differentiation across four miners, and test case rotation across three BOPTEST scenarios. Weights set on-chain via Yuma consensus.

  • intoclaw
  • Shell
  • ★ 4
  • 2026-04-08

intoclaw

A skill pack that teaches Claude-based agents to work in the Bittensor ecosystem.

IntoClaw gives any OpenClaw agent real knowledge of the network, live on-chain data through TaoStats, decentralised web and X search through the Desearch subnet, and a skill-creator that enforces the pack's own structure. The README is written for the bot, not the human: the agent walks the user through what it just picked up, installs dependencies with narration, and never touches a wallet without explicit confirmation.

  • subnet-valuator
  • Python
  • ★ 0
  • 2026-02-04

subnet-valuator

Automated valuation engine for Bittensor subnets. Scores 0-100 derived from on-chain data, Grok research, and Claude scoring.

The pipeline runs raw on-chain data through a Grok research pass, a Claude scoring pass, and a deterministic engine that weights mainstream potential against a fundamental price derived from daily operational expenditure, then applies flow and social stability modifiers. Outputs per-subnet reports and a master ranking. Designed to be re-run end-to-end on a weekly cadence with a single command.

  • repo-racer
  • TypeScript
  • ★ 0
  • 2026-03-07

repo-racer

A real-time 3D racing visualisation driven by live network performance data.

On Bittensor, 30-day net flow sets top speed, 1-day flow sets acceleration, and 7-day flow sets handling. Cars race a parametric circuit with curvature-aware racing lines, overtaking logic, and pit stops. Built with React Three Fiber, a dynamic overhead drone camera, and the IntoTAO design system. An experiment in whether network data can be legible as a race rather than a chart.

no other public repos yet. see github.com/vanlabs-dev

About

Day job is infrastructure and Microsoft cloud work for a New Zealand MSP. M365, Azure, Entra, Intune, client environments across the country. Early mornings, lunch breaks, nights and weekends go into the Bittensor ecosystem: subnet research under IntoTAO, OpenClaw agent tooling, and Zhen, a decentralised calibration layer for digital twins running on testnet. The MSP work keeps the fundamentals sharp. The Bittensor work is where the energy is going.

Talk through the approach first, then build. Most of the time that means a strategy session before touching code, then implementation with clear prompts and tight feedback loops. Methodical over fast. A well-scoped thing beats a rushed one. Systems get a CLAUDE.md before they get a feature. When something fails, it goes into CLOSED_DIRECTIONS so the next experiment doesn't repeat it. The goal is a codebase that a future version of me can actually work in.

Looking to move into a full-time role in the AI and/or Bittensor space where the day job and the side projects are the same thing. Open to building internal AI tooling, working on agent infrastructure, or anything that sits at the edge of practical AI and real systems. Auckland-based, open to remote. If you're building something interesting in decentralised AI or the agent space, it's worth a conversation.

subnet infrastructure · agent tooling · data platforms

Contact