Final-year CS student, graduating 2026. I build full-stack applications with a focus on backend architecture, API design, and systems that are easy to reason about. Currently open to graduate software engineering roles in London.

📍 London, UK · darrelkelyan@outlook.com · LinkedIn

A full-stack music streaming platform with a real-time, event-driven recommendation engine. Built as my Computer Science FYP.

The recommendation engine captures every play, skip, and like as a weighted engagement signal — play: +1, skip: −2, like: +3 — grounded in the implicit feedback model by Hu, Koren & Volinsky (2008). Each interaction triggers a server-side cache invalidation so the next recommendation request reflects fresh behavioural data immediately.

React 18 Node.js Express Firebase Firestore Spotify SDK Render

• Domain-driven custom hook architecture (useTrackEvents, useRecommendations, useSpotifyPlayer) — largest component reduced from ~200 lines of mixed logic to under 50 lines of pure presentation
• BFF OAuth 2.0 flow: backend exchanges the Spotify authorisation code, tokens never touch the client
• Global playback state via React Context (PlayerContext, LikedTracksProvider) — single source of truth across all components
• 63 functional test cases across 9 feature areas; 98% pass rate on first submission

A real-time paper trading platform built with a university peer, combining a Next.js frontend with a Python/FastAPI backend.

Next.js FastAPI Python WebSockets

• Multi-interval stock data API supporting 8 timeframes via both REST and WebSocket endpoints
• Real-time candlestick and line charts with live WebSocket data feeds and automatic candle updates
• Paper trading engine with live P&L tracking, spread calculation, and price-line visualisation
• Asset search with intraday chart previews for discovered symbols

Languages: JavaScript · TypeScript · Python · HTML/CSS
Frontend: React 18 · Next.js
Backend: Node.js · Express · FastAPI · REST · WebSockets
Data & infra: Firebase Firestore · Docker · Git · Postman · Render