I work at the boundary of platform architecture, product, and AI — building the systems and standards that let organisations actually use AI well, not just deploy it. My focus is on how products should be redesigned to let intelligence act, not just assist.
I started as a software engineer, moved through data engineering and data product management, and have most recently been functioning as a technical platform lead and informal TPM — overseeing architectural transitions, setting data standards, and leading AI enablement across engineering teams.
I hold a CISSP and treat audit and event data as a security discipline — not just a data engineering task. My work on event schema governance is designed to answer what happened with integrity and traceability, not just to instrument a product.
I think the most important AI work right now isn't in the models. It's in how systems — and the organisations that build them — are redesigned around intelligence as a primary actor.
Not a CV. Three cases that show how I approach problems at the intersection of systems, product, and AI.
WebMCP is a W3C specification — shipping in Chrome v146+ — that lets AI agents invoke application functionality natively in the browser, without servers or WebSockets. Most product teams don’t know it exists yet. I built two open-source projects to understand where it leads: react-agent-tool, a React hook library that exposes app state and actions as callable agent tools, and webmcp-kitchen-sink, a reference implementation showing enterprise integration patterns. The goal was to get ahead of a paradigm shift before it arrives, not to catch up after.
Most teams treat AI as a passive assistant — something you prompt when stuck. I introduced and operationalised agentic workflows using Claude Code across Kiteworks’ product team, including Git-based development patterns, CI/CD integration for AI-assisted UI development, and MCP-based tool connections. Engineers and PMs now work in an agent-driven way by default, not as an experiment. The result was a structural change in how the team builds — not a productivity feature.
Fragmented event semantics across a distributed product created a system that could instrument activity but couldn’t reliably answer what happened. I designed and institutionalised a company-wide Audit Event Framework — 600+ event definitions governed by JSON Schema data contracts, versioned, auto-documented, and cross-language. I treat this as cybersecurity infrastructure: the goal is integrity and non-repudiation, not just observability. This required rearchitecting the activity data layer, introducing a hierarchical event taxonomy via unsupervised clustering, and performing systematic gap analyses without disrupting product velocity.
Kiteworks is migrating from a Python monolith to a microservices architecture. I own the API gateway layer — the boundary of the system that shapes how services are exposed, secured, and composed. This is not just infrastructure work. How you design the gateway determines what the system can become: which capabilities are composable, which security properties are enforceable at scale, and how external consumers experience the product as the underlying architecture changes. I am also setting the standards by which this transition happens — not only shipping components within it.
On AI systems, product architecture, and where the two are going.
The AI in the app mental model treats models as unreliable components to be contained. That framing is now the ceiling, not the floor. The inversion: stop putting AI inside a workflow — put the workflow inside the AI.
Four positions on AI product strategy held with high conviction: why embedded chatbots are already losing, why architecture should favour adaptability over current capability, and why the locus of decision-making is shifting from explicit orchestration to model judgment.
A new W3C spec lets AI agents invoke application functionality natively in the browser. Most teams haven't noticed yet. This is what it means architecturally — and why I built a reference implementation to find out before the wave arrives.
I'm interested in roles at AI-native companies or AI-forward departments where platform, product, and AI systems intersect. I'm based in Amsterdam and open to relocating across Europe. For the right role, also open to remote with US-based teams.
AI Platform Lead · Technical PM · Head of AI Enablement · Platform Product Manager
Amsterdam-based. Willing to relocate in Europe.
Singaporean — eligible for US H1B1 visa.