03V1.3.14 · GOOGLE PLAY
Founder & full-stack engineer
Small shops with patchy internet need a till that never stops. I built XPOS, a Flutter POS that runs fully offline on a local SQLite database and syncs to the cloud in the background. The key decision was splitting the backend into two Supabase projects: one for billing and plans that shop staff can never reach, and one for shop data locked to each tenant with Postgres Row Level Security. Activation, billing webhooks and AI invoice scanning run in Supabase Edge Functions, with Lemon Squeezy handling subscriptions.
- Flutter (Dart)
- Supabase (Postgres, Auth)
- Supabase Edge Functions
- Drift (SQLite)
- Riverpod
- Lemon Squeezy
- Google Gemini
- GitHub Actions + Vercel
04LIVE · GOOGLE PLAY
Solo developer and designer
Most wallpaper apps are slow, cluttered and hard to keep fresh. I create the art with Nano Banana and built WallGet in Flutter so I can publish it just by uploading it to a folder on ImageKit, with no app update and no custom backend. The app reads folders as categories and uses ImageKit URL transformations to load a small thumbnail in the grid, a sharper preview on the detail screen and a full-size file for the wallpaper. I moved from Supabase to this CDN-first setup to cut moving parts. Bloc manages state, a Kotlin platform channel sets the wallpaper natively, and AdMob with GDPR consent handles ads.
- Flutter (Dart)
- Bloc state management
- ImageKit (CDN, Media API)
- Kotlin platform channel
- Android WallpaperManager
- Google AdMob
- Google UMP consent (GDPR)
- Play In-App Review
05LIVE · GOOGLE PLAY · APP STORE
Flutter engineer, 10-person team
Home health clinicians lose hours typing visit notes after each patient. AutoMynd records the visit with live Azure Speech-to-Text and turns it into OASIS-E1 and follow-up notes the clinician reviews. In a 10-person team at a Sri Lankan partner company, I built the Med-Recon medication flow, auto-correction for live transcripts, and the mapping that uses OpenAI to place extracted answers into the right survey questions. I also handled encrypted, HIPAA-aligned data storage in the app, fixed UI bugs and published releases to Google Play and the App Store.
- Flutter (Dart)
- Azure Speech-to-Text
- OpenAI API
- Azure SQL
- Firebase
- Android & iOS
- CI/CD (GitHub)
- Figma
06PUBLISHED 2025 · RESEARCHGATE
Lead author & full-stack engineer
Fixed exit signs cannot react when a fire blocks a corridor. I built ExitMatrix, a research system that plans a new escape route the moment the fire map changes. A web editor stores each floor as a grid in Firebase Firestore, a Python service runs a trained Q-learning agent, a Node.js engine adds fire safety buffers and turn-by-turn steps, and a Flutter app speaks each step aloud. I used Firestore realtime listeners as the shared link so every part reacts to the same live map with no polling. A Unity 3D building let me test routes safely.
- Flutter
- Unity (C#)
- Python (NumPy, FastAPI)
- Node.js (Express)
- Firebase Firestore
- Firebase Auth
- Firebase Hosting
- Q-learning (RL)
07R&D · ESP32-CAM + ESP8266
Hardware & full-stack engineer
I built a way to control real devices with nothing but hand motion, covering the full loop from camera to switch. An ESP32-CAM streams video over Wi-Fi to a Python server, where MediaPipe tracks the hand and OpenCV turns it into finger counts and rotation angles. The server drives an ESP8266 that switches LEDs and a buzzer and sets motor speed with PWM, all in real time. To keep it fast, I held one warm HTTP connection open and grouped commands into a single batch request. A Vue dashboard shows the live feed and device states.
- Python
- Flask + Socket.IO
- MediaPipe Hands
- OpenCV
- Vue 3 + Tailwind CSS
- ESP32-CAM (Arduino C++)
- ESP8266 (Arduino C++)
- mDNS + HTTP over Wi-Fi
08IOT PROTOTYPE · 2023
Flutter developer, ML integration
Owners who leave their dog at home cannot tell if it is too hot, stressed or unwell. PawTrack+ pairs a smart cage with a mobile app. Arduino sensors read temperature, humidity and light and drive automatic climate control, while camera and motion data feed TensorFlow models built by a teammate that flag abnormal behavior and obesity. I built the Flutter app and connected it to those models through FastAPI, with Firebase syncing live readings and alerts, plus a pet-book for medical history, diet tips and exercise tracking.
- Flutter (Dart)
- Arduino (C++)
- Firebase
- FastAPI (Python)
- TensorFlow
- Temp, humidity, light
- Motion sensor + camera
09AI + IOT PROTOTYPE
Flutter developer, ML integration
Cucumber growers often spot disease too late and sell without knowing where prices are heading. AgriCuke puts four AI tools in one Flutter app. Teammates built the models: a TensorFlow image model for Belly Rot, Downy Mildew and Powdery Mildew, a land suitability model using soil pH and moisture from Arduino field sensors plus weather, and a market price predictor. I built the Flutter app and connected it to these models through FastAPI, so a farmer can scan a leaf or check land and prices in a few taps.
- Flutter (Dart)
- TensorFlow (image models)
- Python (ML models)
- FastAPI
- Arduino (soil sensors)
- Firebase
- Weather API
10IOS APP · NODE.JS API
iOS and backend developer
Online clothing stores often lose shoppers to clumsy search, browsing and checkout. I designed Vogue Vista in Figma, from low fidelity wireframes to a clickable prototype, then built it as a native SwiftUI app using MVVM. The app talks to a REST API I wrote in Node.js and Express, backed by PostgreSQL and hosted on Heroku with deploys from GitHub. Passwords are hashed with bcrypt and login returns a JSON Web Token. The cart lives on the server, so it follows the user to any device, and the API is documented in a public Postman collection.
- Swift (SwiftUI, MVVM)
- Node.js + Express
- PostgreSQL
- Heroku
- JWT + bcrypt
- Figma
- Postman
11V1.0.1 · ANDROID + IOS
Flutter developer
Flight trackers show where a plane is, but not what the trip was actually like. I built FlyMoon in Flutter to do both. Users search a flight number and get live status, delays, terminals, gates and airport, airline and aircraft details from the AviationStack API, with airports pinned on Google Maps. They then rate the journey across 15 categories, from check-in to baggage claim, and earn points for every answer. Firebase handles Google, Apple and email sign-in, while Firestore and Realtime Database keep each trip session so ratings survive app restarts.
- Flutter (Dart)
- GetX
- AviationStack API
- Firebase Auth
- Cloud Firestore
- Firebase Realtime DB
- Google Maps SDK
- Google + Apple Sign-In
12OPEN SOURCE · GITHUB
Embedded and IoT developer
Gas leaks in a kitchen or storeroom can go unnoticed until it is too late. I built a monitor that watches both rooms at once. An Arduino Uno reads two MQ2 gas sensors and drives an RGB LED and buzzer in each room, with short beeps for a minor leak and a continuous siren for a serious one. An ESP8266 NodeMCU reads two DHT11 sensors, receives gas levels from the Uno over serial, and streams everything to a Blynk cloud dashboard where each room's monitoring can be switched on or off. I used two boards because the ESP8266 has only one analog input.
- Arduino Uno
- ESP8266 NodeMCU
- Blynk IoT (Edgent)
- C++ (Arduino)
- MQ2 gas sensors
- DHT11 sensors
- RGB LEDs + buzzers
13AI MUSIC · RESEARCH APP
Flutter developer, ML integration
Music apps recommend from listening history, but ignore how you feel right now. VibraSync starts from your current state. Cam Mix reads emotion from a selfie, Voice Tunes detects mood and stress from a short voice clip, Age | Gender adds a demographic signal, and 360 Mix blends them with the local weather. A teammate built the TensorFlow models; I built the Flutter app, including the Vibra Player, and connected it to the models through FastAPI. After listening, the app asks how you feel and stores it in Firebase.
- Flutter (Dart)
- TensorFlow
- Python
- FastAPI
- Firebase
- Weather API
14WEB APP · PHP + MYSQL
Full-stack PHP developer
Ordering food by phone is slow for customers and busy for restaurants. I built a multi-restaurant ordering website in PHP and MySQL with a JavaScript front end. Customers register, browse restaurants by category, add dishes to a cart, check out with cash on delivery and follow each order as it moves from in process to closed. A separate admin panel lets owners add and edit restaurants, categories, menus and users, and update every order with a status and remark. The site was deployed on Azure App Service.
- PHP
- MySQL
- JavaScript
- HTML + CSS
- Azure App Service
15REACT SPA · NODE.JS API
Full-stack developer
Sri Lanka's Department of Meteorology updates its public weather data only every three hours. I built Geo-360 Live to show conditions every five minutes. A Node.js and Express API pulls readings for 15 cities from the Weatherstack API into PostgreSQL on a schedule and serves them as JSON with Swagger docs. When the free API tier ran out, I added a data simulator on the same pipeline for the demo. A React app plots each city on a Leaflet map with 24-hour stats and email digests, deployed to Heroku and Azure.
- React
- Leaflet
- Node.js + Express
- PostgreSQL
- Swagger (OpenAPI)
- Nodemailer
- Heroku + Azure
- GitHub Actions
16UI/UX CASE STUDY · BEHANCE
UI/UX designer
Most smartwatch fitness apps cram data onto a tiny screen, so users struggle to read heart rate, sleep and progress at a glance. With my team, I studied competing apps and gathered user feedback to find the biggest pain points. I then designed FitFlow, a clean watch interface with gesture-friendly layouts, live data visualisation and custom goal tracking. The screens were designed and prototyped in Figma and refined in Photoshop, and the full case study is published on Behance.
- Figma
- Adobe Photoshop
- Competitive analysis
- User feedback
- Prototyping
17CONCEPT · UI/UX · BEHANCE
UI/UX designer
Bank of Ceylon's online banking was designed for desktops, while a new generation of Sri Lankan customers now banks mainly on their phones. In this unofficial concept, I reimagined the web app mobile-first, with a focus on accessibility, clarity and a sense of security. I started with user research, turned it into personas and user journeys, then moved from wireframes to a high-fidelity prototype in Figma, with visual assets in Photoshop and Illustrator. The result is a simpler, modern flow for everyday transactions.
- Figma
- Adobe Photoshop
- Adobe Illustrator
- User research
- Personas + user journeys
- Wireframing
- Prototyping
18IOT PROTOTYPE · ESP32
Hardware, firmware & web engineer
Earthing rods work best in moist soil, and dry ground pushes earth resistance up. I built a low-cost device that watches the soil and waters it only when needed. An ESP32 reads a moisture probe and measures real earth resistance in ohms with a two-rod voltage divider, flipping probe polarity each cycle to slow corrosion. A 40 to 60 percent hysteresis band drives a relay and pump without chatter. Readings flow through Firebase Realtime Database to a Chart.js dashboard deployed by GitHub Actions.
- ESP32 (Arduino C++)
- Firebase Realtime Database
- Firebase Authentication
- Firebase Hosting
- Chart.js
- HTML, CSS, JavaScript
- GitHub Actions
- Relay + DC pump hardware
19CLIENT SITE · SRI LANKA
Designer & developer, PowerX Global
Salon Sanjalee is a local bridal salon whose old website was a single hand-written HTML page. I rebuilt it as a premium one-page site in Next.js and React. Since the owner already lives on WhatsApp, I skipped a booking backend: the form writes a ready-to-send WhatsApp message instead. Every page is prerendered as static content on Vercel, so it is fast, cheap to host and has nothing to maintain. Motion and Lenis drive the scroll animations, and schema.org data helps it rank in local search.
- Next.js (App Router)
- React
- TypeScript
- Tailwind CSS
- Motion
- Lenis smooth scroll
- Schema.org JSON-LD
- Vercel
20LIVE · VITAL-CHAIR.WEB.APP
Full-stack & embedded engineer
Caregivers of wheelchair users cannot easily check on them or help them move from a distance. I built Vital Chair, a two-board ESP32 wheelchair with a Flutter app and a web dashboard. One board drives the motors from a joystick, a gesture glove or phone commands, and the other reads pulse, SpO2, temperature and falls. I replaced the original Blynk and Bluetooth design with Firebase Realtime Database, so the chair, app and dashboard never talk directly and each part can be tested alone. Firebase Auth and security rules protect every account.
- ESP32 (Arduino C++)
- Flutter (Android, iOS)
- Firebase Realtime Database
- Firebase Auth & Hosting
- Vanilla JS & Chart.js
- MAX30102, ADXL345, DS18B20
- L298N motor driver
- Speech-to-text
21LIVE · OVVIONSHOP.COM
Front-end engineer
OVVION needed a storefront as loud as its hand-drawn alien artwork, without a heavy backend for a small first drop. I turned the original design into a Next.js App Router site with pixel-level parity, including CRT scan lines, marquees and floating mascots. The key decision was one typed product catalog as the single source of truth: it drives the shop grid, statically generated product pages, social preview cards and the sitemap. The cart runs on Zustand with local storage, built with TypeScript and Tailwind CSS on Vercel.
- Next.js (App Router)
- React
- TypeScript
- Tailwind CSS
- Zustand
- Vercel
22RESEARCH · LOCAL WEB APP
Full-stack & AI integration
Eye disease screening needs specialists, and AI models that only output a label are hard to trust. I built a two-tier system: a Flask web app for upload and results, and a FastAPI service that runs a YOLO detector trained on fundus images. The key decision was to never show a prediction without an explanation, so every scan gets occlusion, LIME and SHAP heatmaps. Gemini writes the care report, and every external call has a local fallback so the demo never breaks.
- Python
- FastAPI
- Flask
- Ultralytics YOLO
- PyTorch
- Roboflow Inference
- Google Gemini API
- LIME, SHAP, OpenCV
23PRE-LAUNCH · SRI LANKA
Founder & full-stack engineer
Couples in Sri Lanka still print cards and chase RSVPs by phone. I built Aradhanawa, a freemium platform with 40 animated invitation designs in English, Sinhala and Tamil, a live editor, guest list import, RSVPs, a seating planner and QR check-in at the venue. The key decision was to enforce every plan rule on the server and make payment fulfilment idempotent, so a webhook retry or an admin approval can never publish twice. It runs on Next.js, Postgres with Drizzle, Better Auth and Vercel Blob.
- Next.js (App Router)
- React + TypeScript
- Tailwind CSS + Motion
- Postgres (Neon) + Drizzle
- Better Auth
- Vercel Blob
- PayHere
- Resend