Skip to content

About

I left finance to build the thing, not report on it.

A decade later that’s still the job: start with the business problem, build the real thing, and stay honest about how it went. I build agentic AI, make it adoptable, and ship the scalable systems it runs on. Proof in capital markets, legal AI, and scale-ups. Here’s the longer version.

Editorial portrait (4:5 · 800×1000, duotone)
Reserved — photo to follow

Where it started

I started in finance, at John Keells Holdings. It was a fine place to learn how a business actually keeps score, but I kept wanting to build the systems we reported on instead of reporting on them. So I left, learned to ship, and never looked back.

Architecting a product suite

At TWC Innovations I went from intern to the de-facto architect in four years. Not by waiting my turn, but because the work landed. I designed and built a modular field-sales product suite (sales-force automation, a CRM, a CMS, a client portal) deployed across Asia. I rebuilt the mobile apps as a single cross-platform codebase to open the iOS market, replaced the legacy PHP backend with an API-first one, stood up CI/CD, and trained the engineers who came in after me. It’s where I learned the thing I still lead with: start with the workflow you’re replacing, not the framework you’re using.

Leading at scale

At :Different, a venture-backed proptech that raised $35M+, I led a small engineering team that owned all of communications: email, SMS, push, and in-app, across a multi-tenant white-label platform. I came in to modernize battle-tested code, not to greenfield it. I learned to keep a team shipping while rebuilding the foundations under it, and to translate cleanly between product, stakeholders, and engineers.

Going deeper on AI

Then I went deeper on the AI side: an MSc in Big Data Analytics, and published research on a RAG and prompt-engineering pipeline for legal documents, built and evaluated for about $30, across 1,200+ Sri Lankan court judgments and an 89-category taxonomy. I wrote the paper with a frank retrospective on its own limits, including where my evaluation was biased, because pretending the numbers were cleaner than they were would have been the easy, wrong thing to do.

Consulting

Most recently I’ve consulted on agentic AI for capital-markets infrastructure: autonomous workflows built so a bank’s security and procurement teams would actually deploy them: enterprise SSO, audited and explainable output, integrations that hold on the third retry. The through-line across all of it: the model is rarely the hard part. Making the thing adoptable is.

What I care about

Code that works isn’t enough; it has to work for the business. I’d rather ship something honest and adoptable than something impressive in a demo. And I’m wary of engineering that ignores the people who actually use it. That’s usually the part that decides whether the thing survives contact with production.

The trajectory

Four engineering rungs, one habit.

  1. WorkWave / IFS , Lead Engineer

    2024 →

    ServiceBot: first from the Sri Lankan team on it. UX improvements, ~30x performance gain, Service Express (v2 product), four WorkWave suite integrations, notification fan-out, TypeScript migration, code-level security work, and a WordPress plugin.

  2. Karmo , Senior Engineer

    2023–24

    Subscription billing, automated credit checks, inventory, and the customer portal for a car-subscription platform: 200+ vehicles across 35+ brands.

  3. :Different , Senior Engineer + Team Lead

    2022–23

    Led a small engineering team owning multi-tenant comms (email, SMS, push, in-app) at a $35M-funded proptech.

  4. TWC Innovations , Intern → Engineer → Architect

    2018–22

    Designed and shipped a modular product suite across Asia: cross-platform mobile (one codebase, iOS + Android), an API-first backend rebuild, CI/CD, and mentoring.

  5. John Keells Holdings , Finance

    2016–17

    Where it started, and where I decided to build the systems instead of reporting on them.

Off the clock

Curiosity over comfort.

I play games to understand how good systems are designed — the kind of craft that only becomes visible when something is built with genuine care. The same instinct shows up in the work: I want to understand the whole system, not just the part I was handed.

Working on something that has to ship for real?

Tell me the problem See the work