Luke Enness

Education

Swansea University

BSc (Hons) Computer Science with a Foundation Year

Technical skills

• Languages: TypeScript, Python (3.14+), SQL, Java, HTML5/CSS3
• AI / ML & Simulation: PyTorch, Optuna, Reinforcement Learning, Kaggle Environments
• Frontend: React 19, Vite, TanStack Router, TanStack Query, Tailwind CSS v4, Three.js, Lucide React, Google Maps Platform
• Backend & Databases: FastAPI, Pydantic, PostgreSQL (Supabase), Docker Compose, uv, PyArrow
• DevOps, Testing & Quality: Git, GitHub Actions CI/CD (Vercel, Render), Pytest, Ruff, ESLint

Projects

playsouthwales.uk  — Tournament Pairings & League Scheduling SaaS FastAPI, React 19, TypeScript, Supabase, Tailwind CSS v4

• Developed and launched a tournament pairings and league scheduling web app for the South Wales competitive Pokémon community designed to solve player drop-off and scheduling clashes.
• Built a responsive frontend using React 19, TypeScript, TanStack Router, TanStack Query, Clerk Auth, and Tailwind CSS v4 backed by a FastAPI backend with a Supabase PostgreSQL database and Pydantic models.
• Wrote async background workers to automate the collection of relevant tournament data such as major events and release dates from community APIs and websites.
• Set up Docker Compose for local development and automated CI/CD pipelines for deployments to Vercel and Render.
• Built a companion Discord bot that automatically posts tournament pairings, schedules, and standings directly to local league servers.
• Scaled platform to 70 weekly active users and 200+ page views, onboarding 16 tournament organisers across South Wales.

Jeremy — Autonomous Agent & Simulation Framework Python, PyTorch, Optuna, Pytest, NumPy

• Engineered an autonomous agent framework for the Kaggle Kaggriculture competition, tested with a 68-test Pytest suite.
• Implemented a spatial dispatcher that coordinates farm units across a grid by scoring chore priorities and Manhattan distances each turn.
• Designed a NumPy-based neural macro-policy for daily farm management tasks trained from supervised learning datasets generated from Parquet replay logging scripts.
• Ran parallel Optuna studies to tune multi-agent action priorities such as planting, watering, and harvesting.
• Automated bundling, verification, and performance benchmark scripts for submitting agents directly to Kaggle tournaments.

Stunfisk — Reinforcement Learning PTCG Agent Python, PyTorch, MCTS, Reinforcement Learning

• Built an AlphaZero-style reinforcement learning agent combining neural network evaluation with Monte Carlo Tree Search (MCTS) to play competitive Pokémon TCG.
• Trained value networks using Generalized Advantage Estimation (GAE) and HL-Gauss distributions for turn evaluations while optimizing PyTorch training loops to run efficiently on local hardware.

Spectrophotometer — Laboratory Data Analysis Tool TypeScript, Vite, HTML5 Canvas API

• Created a browser-based analysis tool in TypeScript and Vite, replacing legacy proprietary Windows XP software for university researchers.
• Created a client-side parser that takes raw .txt measurement files and plots interactive spectral response curves on HTML5 Canvas.
• Currently used by Swansea University academic faculty and students to process data for dissertation research.

CTHead Scan — Visual Computing Project JavaFX

• Engineered a desktop application to process and render 3D medical CT scan volumes.
• Developed the core rendering algorithms Maximum Intensity Projection (MIP) and 3D direct volume raycasting in JavaFX and presented a live viva to university lecturers achieveing 100% on the project.