Computer Science undergraduate at the University of Cambridge (Churchill College). I like utilising my software engineering knowledge, and have a keen interest in low-level programming.
Selected from around 1,000 applicants for Google's one-day AI hackathon in London, building alongside Google engineers and pitching to a panel of their judges. Built a voice-first crisis reporting app on Gemini Live, with barge-in driven by server-side voice activity detection and structured incidents written to a live responder feed.
Won a $2,500 prize placing 2nd in an invite-only, week-long algorithmic game theory competition. Built an adaptive attack-and-defence Python strategy with a team collaborating across Europe via Git.
Conceptualised and pitched a proposal using Visa's API infrastructure to integrate an AI prediction model for real-time credit-score impact visualisations within banking apps.
Administered Windows Server environments for secure cross-branch data exchange; configured a Squid web proxy on Ubuntu Server and cut cached-content retrieval times by 60% through server-side caching refinements.
Limit order book and matching engine in C++20 with price–time priority and five order types. O(1) cancels via cached list iterators, upfront FillOrKill feasibility checks using incremental per-level quantity aggregates, and thread-safe GoodForDay expiry via a background prune thread. Validated by a data-driven GoogleTest scenario suite and profiled with a latency microbenchmark.
Voice-first crisis reporting, built at Google's Student AI Hackathon. A React front end streams 16kHz microphone audio over a WebSocket to a FastAPI backend that owns the Gemini Live session, with barge-in over server-side voice activity detection and an idempotent tool call keeping duplicate reports out of SQLite. Responders triage incidents live on an interactive 3D globe with urgency-based colour coding.
Leakage-safe, isotonic-calibrated win-probability model fusing FACEIT Elo with recent-form stats - less than half the miscalibration of the raw Elo baseline (ECE 0.07 → 0.03). Benchmarked on 5,000 real matches with paired-bootstrap CIs, DeLong tests and a 4,400+ prediction walk-forward backtest, fed by a checkpointed snowball crawler and shipped as a live cached Django app.
RAG-based revision agent that retrieves past exam questions via semantic search. Powered by a multithreaded computer-vision + ETL pipeline that ingests hundreds of PDF pages while preserving LaTeX, cutting processing time by 75%.
Modules include Foundations of CS (OCaml), OOP (Java), Software & Security Engineering, Databases, Algorithms I & II, Machine Learning & Real-World Data, Operating Systems, and Digital Electronics.
Outside of lectures I do many things, from playing college football and competitive programming all the way to chess!
Mathematics, Further Mathematics, Physics & Chemistry. UKMT SMC Gold and Senior Kangaroo (#1 in college for both), Smailes Goldie Award for Best Application of Mathematics, BPhO Computational Challenge.
If you'd like to chat - about a role, a project, or anything else - my inbox is always open!