G3 — a family health app, built local-first
SwiftUI and SwiftData on iOS, with no server and no account. An exercise in how much discipline a one-person app can carry.
Senior Director, Data, ML & GenAI Platform · GoDaddy · Greater Seattle
I build the platforms enterprise AI runs on.
Twenty years in platform engineering and engineering leadership — ten of them at Amazon, where I founded an organisation and built systems that searched ten billion items. Today I lead company-wide GenAI platform architecture at GoDaddy: agent runtimes, evaluation frameworks, and the integration layer underneath them.
- 20 yrs
- Platform engineering and engineering leadership
- 80+
- Largest engineering organisation led
- 10B
- Catalog items indexed for search, across 16 countries
- ~90%
- Code generation in an AI-first SDLC
Selected work
Systems I designed, led or built, and what they ran at.
-
An enterprise GenAI platform
Company-wide architecture and strategy for LLM-based and agentic systems, behind both customer-facing products and internal platforms. Distributed LLM infrastructure, RAG pipelines, long- and short-term agent memory, context orchestration and personalisation. A unified agent platform — SDK, runtime and UX layer — so product teams ship agents without each rebuilding the same primitives, plus MCP-based connectors giving them one standard contract into enterprise systems.
- 30% faster AI time-to-market
- ~90% code generation
- $MM cost savings
-
Evaluation and observability for agents
The part that decides whether any of the above can be trusted in production: evaluation, observability and instrumentation frameworks for GenAI systems, so behaviour is measurable, regressions are visible, and governance is something the platform provides rather than something each team improvises.
- reliability
- performance
- governance at scale
-
Brand protection at catalog scale
Founded and grew the Automated Brand Protection organisation from nothing to 40+ people, and built the detection and enforcement systems behind Brand Registry and Project Zero. A rule engine for catalog infringement in text and images; an Elasticsearch index spanning the whole catalog; and visual image search using machine learning to detect trademark infringement. Optimising the rule engine onto AWS Elasticsearch cut cost fivefold with zero downtime — patent filed.
- 10B items · 16 countries
- 10k transactions/sec
- 50k+ rules
- 3M images/day
- 100k+ logos · 99% precision
- 5x cost reduction
-
Generative AI for Smart Vehicles
Senior technical owner for LLM initiatives across the Smart Vehicles cloud domain — car control, personal mobility and navigation, and automotive services including EV charging.
- LLM initiatives
- cloud services ownership
Career
Singapore, then Seattle. The longer version →
- 2024 – now
Senior Director, Data, ML & GenAI Platform · GoDaddy
Company-wide GenAI platform. Earlier in the role, data platform and ML systems.
- 2014 – 2024
Senior Manager, Engineering & Science · Amazon
Smart Vehicles generative AI; before that brand protection and trust & safety, 80+ engineers, TPMs and applied scientists across 13 locales and 20 languages.
- 2011 – 2014
Software Development Manager · Expedia
Vacation Packages backend across four teams. A rule-based pricing engine replacing a rigid legacy system, MongoDB and Hadoop, and quarterly releases cut to three weeks.
- 2005 – 2011
Software Engineer · Microsoft
Subscription and commerce platform; Bing Rewards and advertising.
- 2004 – 2005
Software Engineer · ANTLabs, Singapore
Built by hand
Side projects. At this point most people have stopped building; I haven't.
A small serverless platform, and its sharp edges
React, FastAPI on Lambda, and a single-table DynamoDB store running a handful of private apps. Mostly interesting for the things that were hard to debug.
Contact
Greater Seattle.
Email is fastest, or a LinkedIn connection with a line about what you’re working on.