CS @ University of Colombo School of Computing '26 · Building production systems
Currently: ERP Consulting @ EY · COO @ JAK Travels · Shipping Brux Waffles & RideLink

Full-stack engineer focused on shipping real systems for real users. I've built a full-stack e-commerce platform for a paying client, an AI-powered desktop tool that cuts API costs by ~80%, and I'm currently building a carpooling startup. Comfortable across the stack — TypeScript/Node on the backend, Next.js and React Native on the frontend, with a strong interest in backend architecture, payments, and AI integrations.

Outside of code, I'm a competitive boxer and Vice Captain of the UCSC Boxing Team.

Café ordering platform built for a paying client in Colombo. End-to-end flow: menu, cart, checkout, payment, live order tracking, kitchen display, and admin dashboard. Frontend deployed, full launch in progress.

• Stack: Next.js 14, Medusa v2, Supabase, PostgreSQL, PayHere, Google Maps, Vercel, Railway
• Highlights: Server-side price validation · Zone-based delivery fees · Custom waffle builder · Kitchen display system · Full security audit (httpOnly cookies, HSTS/CSP, rotated keys) · Resend transactional email + Medusa event-bus password reset

Desktop application that tailors CVs for specific job descriptions using a two-stage AI pipeline. Engineered prompt caching to cut per-CV token cost by ~80%, with programmatic guardrails that strip AI hallucinations before they reach the document.

• Stack: Python, PyQt5, Anthropic Claude API, python-docx, pdfplumber
• Highlights: Two-stage AI pipeline (fit assessment → generation) · ~80% token cost reduction via prompt caching · Hallucination validation against a source-of-truth profile · Bulk mode with multi-threaded API calls

Mobile-first carpooling platform with driver/passenger flows, real-time updates, and route alternatives. Currently building toward launch.

• Stack: Expo, React Native, Redux Toolkit, Express, Node.js, Firebase Admin SDK, Firestore, Google Maps
• Highlights: Alternative route selection with encoded polylines · Fuel-type-aware fare calculation · AsyncStorage autocomplete caching (7-day TTL) · 2-free-rides/month freemium model

End-to-end automation pulling hotel rate emails from Outlook via Microsoft Graph API, OCR-extracting PDF/image attachments, and structuring data via Gemini LLM function calling.

• Stack: Python, FastAPI, Microsoft Graph API, Tesseract OCR, Google Gemini, Airtable
• Highlights: Natural-language chat interface for non-technical staff · LLM function calling for structured extraction · Production deployment serving JAK Travels operations

Multi-device digital scoring system replacing manual scorecards in boxing competitions. PIN-based role auth (admin/judge/display), live score sync via Firestore.

• Stack: TypeScript, React, Firebase, Firestore
• Highlights: Real-time multi-judge sync via onSnapshot · Three role-based UIs · Mobile-optimized for ringside use

Languages

Frontend

Backend

Databases & Cloud

AI & Integrations

• Shipping Brux Waffles to full launch (Phase 2 PickMe integration in scope)
• Pushing RideLink toward public launch
• Targeting graduate / early-career SWE roles - open to remote and Singapore-based opportunities

Open to interesting backend, full-stack, and AI-integration work. Reach out on LinkedIn.