Applied AI Engineer · United Kingdom
I build systems the people relying on them can inspect, question and trust.
End-to-end ML systems: data pipelines, models, authenticated APIs, AWS deployment, live monitoring.
01 / Projects
Secure data & monitoring platform
Real-time events ingested into PostgreSQL through an authenticated FastAPI service on AWS (EC2, RDS, Docker). Live Streamlit/Plotly dashboard with human-confirmed alerting, plus a React Native companion app with push notifications.
LLM decision-support app
Node.js app built on Anthropic's Claude Agent SDK. Reads freelance job postings from pasted text or screenshots (multimodal), flags each one APPLY / CAUTION / AVOID against user-defined rules, and streams structured JSON. Every draft passes human review.
94.3% → 69.1%
Reported accuracy → verified accuracy, after finding and eliminating data leakage. Published as a public correction, not a quiet fix.
I found data leakage in my own published results, and said so. Rebuilt the dataset from frame-level ground truth, moved to event-level metrics (detection rate, alarm latency, false-alarm rate), ran SHAP explainability audits, and shipped two models with a documented sensitivity trade-off.
The flattering number was easier to publish. The honest one is the reason you can trust the next one.
02 / Writing & community
Towards Data Science
Artificial Intelligence in Plain English · 39K+ followers
MLOps Community · 47K+ followers
03 / Experience & education
May 2024 to present
Designed, built and deployed a responsive React website for Science Behind Hairdressing, then evolved it across three releases as the business changed, integrating Stripe, Fresha and Kajabi for a non-technical stakeholder.
2022 to 2023
Defect-pattern analysis and training colleagues on digital record-keeping.
MSc · Merit
Dissertation: pre-impact fall detection from wearable accelerometer data.
04 / Skills
TensorFlow/Keras · scikit-learn · time-series & sensor data · feature engineering
Leakage detection · event-level metrics · SHAP audits · ground-truth rebuilds · sensitivity trade-offs
Python · TypeScript / Node.js · FastAPI · React Native · automated testing & CI
AWS EC2 / RDS · Docker · CI/CD · authenticated APIs · live monitoring & alerting
Claude Agent SDK · multimodal inputs · structured JSON output · streaming · human-in-the-loop review
PostgreSQL · real-time ingestion · Streamlit · Plotly
Open to applied AI / ML engineering roles across the UK. On-site, hybrid or remote.