Associate Lead Software Engineer · Colombo, Sri Lanka

Sashika
Suraweera

I design the patterns teams build on.
Frontend architecture for enterprise SaaS, from 10M-account billing to AI-powered HR.
  • 6+ yrsbuilding for US enterprise clients
  • ~10Mbilling accounts on the platform I architected for
  • 5domains: billing, healthcare, HR & more
  • 19Kweekly npm downloads, open source
  • 15articles published on Medium
photo · unisci24.com
Selected architecture work

Decisions that teams now build on

Each one: the problem, the constraint, the trade-off I chose, and what happened after.

SaaS billing · enterprise clients

Privilege injection for a 10M-account platform

Problem
Permission checks duplicated per route and drilled through props.
Decision
One reusable injection pattern any component can consume.
95+ componentsadopted app-wide
Read the article ↗
SaaS billing · cross-team

Live-update localization without redeploys

Problem
Every label change needed a full release.
Decision
Designed a new REST contract + BFF path; pitched it to the client and the API team.
Contract acceptedrolled out by engineering
Read the article ↗
@blocks (npm)Admin editorPublished site~30 blocks · draft → preview → publish
AI-powered HR SaaS · team I lead

No-code Site Builder

Problem
Editor and public site drifting apart as blocks grew.
Decision
Ship blocks as one npm package both sides consume; MongoDB model for drafts & publishing.
~30 blocksone source of truth
Write-up soon

All work →

How I work with AI

Agents write code. I own the system.

AI makes a team faster only when the guardrails are engineered first. This is what that looks like on my teams.

01

Guardrails before generation

Lint rules and CLI tooling steer coding agents to our design system, so output is on-brand by default.

02

Tests are the contract

160+ test files mean agent output gets proven, not trusted.

03

Every diff gets a human

I review and approve the PRs. Accountability doesn't get delegated.

04

Context is engineered

Skills, prompts and token budgets treated like code, and written up publicly.

“The architecture decisions are still mine. AI just shortens the distance between the decision and the shipped code.”
Built for platforms used by

Enterprise customers across billing, healthcare & HR

Via client engagements at Altrium and Champ IT Solutions.

ComcastVMwarePhilips HealthcareExperianSubaruFalck3MSentiva
Now
updated

What I'm focused on

  • Leading a no-code site builder for an AI-powered HR platform
  • Owning the company design system architecture
  • Writing up the decisions behind it
More on /now →
On my headphones
What I'm listening toOpen my Spotify
Latest release
postal-code-checkerv2.3.0
Latest article
Changing UI Labels Without a Redeploy: A CDN Architecture for Localization
In Colombo
--:--GMT+5:30