Fourteen years of writing software. Lately: cloud architecture, event-driven services, and AI tooling.
Started on front-end work in Kathmandu in 2012, in Bangkok since 2015. I look after a team these days and I am still in the codebase most weeks.
I build software for a fashion e-commerce company in Bangkok, mostly at the platform layer: system design, cloud architecture, event-driven services, database design, and lately AI tooling.
I started on front-end work in Kathmandu in 2012 and moved to Bangkok in 2015. Over time the problems I found interesting moved down the stack — interfaces, then services, then the infrastructure, data and delivery pipelines everything else sits on. Most recently that has meant building MCP servers and agent tooling so the team can put language models to work on real internal problems.
I also look after a team, which means direction, hiring and helping people do their best work. But I try not to make decisions about layers I have not worked in — which is why I am still picking up tickets. It keeps me honest about how a system actually behaves rather than how it looks in a diagram.
Taking a vague requirement to a design that survives contact with production: service boundaries, event-driven flows, API contracts, and the data model underneath. Usually the highest-leverage hour of the whole project.
AWS architecture written as Terraform — versioned, reviewable and boring to operate. Container platforms, managed databases, networking, secrets, access control and the alarms that tell you before a customer does.
Building MCP servers and agent tooling that connect language models to real internal systems, with the boring parts handled: scoped access, predictable context, and a way to tell whether the output is actually any good.
Schema design, query and index work, and moving data between systems without losing any of it. Backups you have actually restored from, because an untested backup is a rumour.
CI/CD pipelines and automated guardrails that shorten the path from commit to production without making it riskier — plus the hiring, review habits and mentoring that keep it that way.
Companies and detail on LinkedIn.
Email works best. I read everything that comes in.