Comp Sci Graduate · Open to Work
A Computer science graduate focused on applied AI engineering. I build AI systems with an emphasis on evaluation, explainability, and reliability, not just getting a model to work, but understanding where it fails and choosing the right architecture for the problem.
My projects span fraud detection, document intelligence, and AI-powered products, combining machine learning with the software engineering needed to turn prototypes into usable systems.
Work
Trained XGBoost on 1.3M synthetic transactions, deliberately optimizing for recall (0.93) over precision, the right tradeoff when missing fraud costs more than a false alarm. Found and fixed three real data quality issues in the source dataset before trusting the model's output. Extended with a RAG pipeline using SHAP values as the retrieval query to pull relevant fraud policy, and an agent that checks merchant history against policy before recommending escalate or clear, with a full audit trail on every decision.
Upload a legal document and ask questions grounded strictly in its own text. Built two complete Q&A pipelines, full-context and RAG, and evaluated them head to head instead of assuming RAG was right by default. Found a real accuracy gap on long documents, traced it to three separate causes, and shipped the evidence-based recommendation over the trendier architecture.
A study tool that reads your actual lecture material before it teaches you. Upload course PDFs, get grounded explanations, get quizzed, and get targeted follow-up questions on exactly what you got wrong. Built end-to-end: FastAPI backend, React frontend, deployed and password-protected to manage hosting costs as a live demo.
Live demo — password required, contact me for access
React Native mobile app built on top of a published RCT (Vowels et al., 2025, PLOS Mental Health), working inside an existing academic codebase rather than a greenfield build, coordinating changes with the research team via Git. Responsible for real-time voice and text chat via Socket.IO streaming, plus a rule-based crisis detection system that flags at-risk users across both input modes, a safety-critical feature with zero tolerance for false negatives. SUS usability score: 75.8 across 7 participants.
Full-stack job application tracker, built after my own tracking notes became an unsearchable mess. Logs every application with status, dates, and a saved copy of the job description so it survives the listing being taken down. Per-user data isolation, rate-limited auth, Postgres-backed sessions. Deployed and used by friends tracking their own job search.
End-to-end cloud pipeline built with a university team of four. Owned the architecture framing and led the narration for the final assessed presentation, translating a multi-stage ingestion-to-visualization pipeline into a story a non-technical audience could follow.
Analyzed simulated Google campaign data to inform a $10M multi-market Gemini Pro student campaign. Ran proportion z-tests across 48 brand lift studies, filtering to 40 statistically significant results, then calculated Cost Per Lifted User by joining validated lift data to historic spend.
Full-stack app built with a team of four, containerized with Docker and shipped via GitHub Actions CI/CD. First real experience with team software workflows: version control discipline, pull requests, and delivering under a shared deadline.
Hands-on labs in vulnerability assessment and penetration testing fundamentals: NMAP scanning, SSH security analysis, and exploiting/mitigating vulnerabilities in OWASP Juice Shop.
Experience
Education
Relevant modules: Artificial Intelligence, Machine Learning, Data Science, Data Engineering, Software Engineering, Cyber Security, Algorithms, Databases.
Extra
Completed the Data Science route of Google's Step Up Career Challenge, analysing campaign data to inform a $10M student marketing strategy for Gemini Pro.
Attended the national Cyber Hackathon hosted by the National Cyber Force — my first CTF. Worked in a team to complete a bespoke Capture the Flag challenge, applying problem-solving and adaptability under real time pressure alongside participants from defence, academia, and industry.
Technical