CORTEX // ONLINE
INTERFACE 01

Hello there — I'm Vinidu

I build the
whole stack,
from the screen
to the silicon.

Full-stack software people actually use, and the AI around it — agents, harnesses, and the orchestration most products never get to. I understand it the whole way down, to the systems most apps just sit on top of. Give me the hardest problem in your stack, the one that has to hold, and I'll own it end to end.

role software · ailoop agentscore full-stacksys cortex
01path

Experience / Education

02selected work

Things people actually use.

Markets, IoT, and a few interaction studies — the same discipline I now put under software and agents. Scroll sideways. Each case is a system that had to hold.

$ ls ./work --horizontal

CASE 01 // SOMA.FINANCElive
SOMA.finance
01

Backend Engineer

2022 — Present

SOMA.finance

Compliant multi-asset DEX

The first compliant issuance and trading stack for tokenized equities. I owned the server-side architecture — the path from a signed order to a searchable book — and the official site that has to look like a market, not a demo.

  • Serverless trading APIs on Lambda, API Gateway, and SQS
  • DynamoDB as source of truth, synced live into OpenSearch for sort and pagination
  • Auth0, Sentry, CloudFront, and the Next.js surface investors actually land on
Next.jsAWS LambdaDynamoDBOpenSearchSQSAuth0
CASE 02 // NEOMAshipped
Neoma
02

Only engineer in HK

2019 — 2022

Neoma

IoT platform & product sites

End-to-end product work as the only developer in Hong Kong — RFID gates, a cluster of field devices, analytics, and the marketing sites travellers hit first. Hardware in the building, software in the cloud, one person holding both.

  • ~100 IoT devices streaming every five minutes through Kinesis Firehose
  • Redshift warehouse plus hourly Lambda analysis into DynamoDB
  • RFID infrastructure, React surfaces, and the public Neoma site
AWSKinesisRedshiftIoT CoreReact
CASE 03 // ON THE LISTshipped
On The List
03

Engineering

Neoma

On The List

Members-only luxury marketplace

A private flash-sale destination for a high-end retail audience. I worked the product surfaces that have to feel exclusive without getting in the way of a drop — browse, campaign, and the moment someone decides to buy.

  • Member-facing catalog and campaign pages for timed luxury drops
  • Built for a brand that sells scarcity, not noise
  • Shipped inside Neoma’s Hong Kong engineering stack
E-commerceCampaignsWeb
CASE 04 // GAMUstudy
Gamu
04

Full Stack

Personal

Gamu

Independent product on the web

An independent web project used to pressure-test how a brand should feel in the browser — pacing, type, and the small product decisions you only notice when you have to ship the whole thing alone.

  • One-person build from concept to a live URL
  • Interaction and brand treated as the same problem
  • A test bed for the same stack I put under client work
Next.jsBrandProduct
CASE 05 // MOTION STUDIESlab
Motion studies
05

Frontend

Lab

Motion studies

Timing, easing, attention

A playground for the movements that make an interface feel considered. Not decoration — timing, easing, and the gap between a click and a response, isolated so they can be reused in real products.

  • Orchestrated sequences instead of one-off tweens
  • Easing curves tuned for UI, not demo reels
  • Patterns that later showed up in production pages
Framer MotionEasingInteraction
CASE 06 // FADE TEXTlab
Fade text
06

Frontend

Lab

Fade text

Scroll typography

Typography that appears and dissolves as you move through the page. A study in attention: when a line should arrive, when it should leave, and how much of the story the scroll itself is allowed to tell.

  • Scroll-linked opacity and pacing for long copy
  • Type treated as a timeline, not a static block
  • A small experiment with an outsized effect on reading
ScrollTypeAttention
//toolkit

Tools I reach for

A working set for software and the AI on top of it — TypeScript and AWS on the server, React when it needs a face.

03about

About me

Hello — I'm Vinidu. I started in electronics, moved to software when a React course beat 5-stage pipelining, and now I build both: full-stack products and the AI systems that sit inside them. Harnesses that keep models honest, agents that use tools, and the software those loops have to live in.

I've shipped at a Web3 pioneer, a start-up, and a bank. That path is how I still work — screen to silicon, interface to infra. Give me the piece of the stack that has to hold. I'll own it end to end.

Portrait
04transmit

Have a problem that has to hold?

Software, agents, or the stack underneath — if it's the hard part, I want it. Reach me on LinkedIn or GitHub.