I build AI systems that make decisions at scale. 10+ years across enterprise, government, and startups. Pega CDH, adaptive decisioning, production ML, and GenAI integration. PhD in Bayesian modelling. Looking for remote outside IR35 contract engagements.
I work at the intersection of data science, product, and engineering. The common thread is taking analytical models and turning them into systems that make real decisions in production.
Adaptive decisioning systems for customer lifecycle management, campaign optimisation, and collections. Hands-on Pega Customer Decision Hub implementation at national scale. Thompson Sampling, Bayesian methods, and self-learning systems that improve with every decision.
End-to-end model lifecycle: training pipelines, scoring, deployment, monitoring, drift detection, and retraining. Recommendation engines, anomaly detection, personalisation at scale. Governance and audit trails for regulated environments.
RAG architectures, embedding pipelines, prompt orchestration, and LLM API integration in production applications with real users. Confidence scoring, evaluation frameworks, and audit controls for regulated environments.
Production systems with real users and measurable outcomes.
Pega Customer Decision Hub implementation for HMRC's Analytically Driven Campaign (ADC) programme. Adaptive model simulation, synthetic data pipelines, and decisioning strategy design for debt collection at national scale.
Built end-to-end as technical founder. FastAPI backend, Claude API for campaign generation, dual scoring engine with Thompson Sampling optimisation, five-tier booking attribution system.
Recommendation system built and productionised. Automated training, scoring, deployment with full governance and audit trail in regulated commercial banking.
Balance and transaction behaviour anomaly detection. Feature engineering, model training, inference monitoring, and shadow testing across commercial banking portfolio.
Offer ranking and customer decisioning at scale. ML pipeline from ingestion through to low-latency inference at 100-150ms. Automated retraining with MLflow.
Public-facing free tool for STR operators to upload guest data and receive scored revenue opportunities. CSV parsing, automated segmentation, and instant reporting.
Available for remote contract engagements, outside IR35. Based in Birmingham, UK.