I build hard AI systems end-to-end and I go deep across the whole stack rather than staying in one lane: NLP and applied ML, controlled generation and retrieval, real-time 3D runtimes, and the cloud infrastructure underneath. I'm at my best early, as a founding engineer or technical co-founder, owning architecture and writing the core myself on a small team. I hold production-grade bars for latency and correctness, and I've done the founder-side work too, so I understand the business, the customer, and the trade-offs, not just the code. Hire me to take an AI product from zero to something real that holds up at scale. My rare edge is pairing real-time character and avatar AI with NLP.
I'm looking to be an early, hands-on technical owner at a startup building real AI: founding engineer, member of technical staff, technical co-founder, or a CTO-track role where I set the architecture and write the core myself. Small teams suit me. I've spent 8 years building a real-time multimodal AI platform end-to-end, NLP through to a live runtime, and I'd rather be four people deep in a hard problem than managing it from a distance.
I'm most drawn to companies where the AI actually has to work in production: tight latency and correctness budgets, real users depending on it, not demos. Applied ML, real-time and multimodal systems, NLP, and anything involving AI-driven avatars or characters are squarely my wheelhouse, and that combination of real-time character AI with NLP is my rare edge. Open to full-time or fractional, whatever fits the stage.
On logistics: I'm a New Zealand citizen relocating to Sydney, so I have automatic, indefinite work rights across both AU and NZ with no sponsorship needed.
Co-founder and CTO
Early-stage AI startup • 2017 - Present
For 8 years I've architected and personally built a real-time multimodal AI platform end-to-end: text and speech in, a live animated avatar out, at ~10s latency and held to production-grade correctness rather than demo-grade. I write across the whole stack: the neural machine translation and controlled-generation layer, a custom intermediate representation and notation framework for consistent output across domains, a hybrid retrieval layer (embedding-based vector search, a semantic knowledge graph, and LLM-based context retrieval) grounding generation against a structured knowledge base, neural avatar rendering using GANs and custom transformers in PyTorch, the real-time 3D runtime in Unreal Engine (C++), and the event-driven AWS infrastructure underneath it all.
Measurable wins: cut animation production from ~1 month per minute of content to 2-3 days, scaled a structured linguistic database past 12,000 entries, and hit near-100% human-rated translation acceptability in critical domains within tight latency and cost budgets.
I also run the technical side of the business. I've led teams up to 7 across AI/ML, software, graphics, linguistics and infrastructure with zero senior attrition, raised $6M+ across four rounds, secured $1M+ in non-dilutive R&D funding, closed $2M+ in enterprise contracts as the technical face in the room, and led ISO 27001 certification and SOC 2 readiness. I'm lead inventor on a granted patent (US, UK, AU) covering the core method.
