Hey there —
Software engineer in Colombo, about ten years in. These days I spend most of my time on AI and machine learning systems and the cloud infrastructure that keeps them standing up — and the rest of it writing and making videos about what I pick up along the way.
I like problems that look messy from the outside and turn out to have a clean shape underneath. Most of what I know came from breaking something first.
I started out as a backend developer who was far too interested in the servers. That curiosity took me through distributed systems, then into machine learning, and eventually into the part nobody really wants to own — the gap between a model that works in a notebook and a system that's still standing at three in the morning.
So that's mostly what I do now. Data pipelines, training and evaluation loops, inference services, and the cloud platforms underneath them. I think a lot about latency, about what things cost to run, and about whether whoever is on call can actually work out what broke. The best systems I've worked on were the boring ones.
The other half is writing and recording. I started doing it because explaining something is the fastest way to find out whether I actually understand it. I kept doing it because the explanations I needed when I was starting out mostly weren't there.
I'm based in Colombo, I work remotely with teams in a few different timezones, and I'm always happy to talk shop with someone who's stuck on something interesting.
A few projects that made it past the demo and into daily use.
Took a single-region monolith to active-active Kubernetes across three regions, with nobody noticing the cutover. All of it in Terraform afterwards, and the monthly cloud bill came down around 40% once we right-sized and moved the batch work to spot.
Nothing in that category yet.
2022 — now
I own the architecture behind the AI products we ship, built the evaluation and inference platform they run on, and work with a team of eight — which has turned out to be as much about writing things down clearly as it is about code.
2019 — 2022
Led the move to Kubernetes and an event-driven core, and introduced infrastructure as code along with an SLO practice that survived a few rounds of healthy scepticism.
2016 — 2019
Shipped features across the whole stack and got properly acquainted with databases, queues, and what a bad schema decision costs you eighteen months later.
2015
A small internal tool, one genuinely nerve-wracking deploy, and the realisation that this was what I wanted to do with my time.
Usually whatever I've just spent a week being confused by.
Working on something interesting, stuck on something annoying, or just want to argue about architecture? I read everything that lands here, and I reply to most of it.
It'll land in my inbox and I'll reply from there. Worth a quick look first.