Seven years building systems that hold under pressure — at Barclays, US Bank, and Amazon, where data quality isn't optional and downtime has a price tag.
At Deloitte I engineered distributed ELT pipelines and analytics platforms for Barclays Investment Banking and Amazon Financial Services — Python on AWS Glue, star schema semantic data models, LLM-powered reporting interfaces. Before most enterprise teams were thinking about generative AI in production, I'd already shipped it.
At Publicis Sapient I designed the GraphQL BFF architecture for US Bank Treasury from scratch — subscription-based WebSocket connections keeping payment state consistent across 50K+ concurrent users. At Lowe's I built checkout infrastructure processing $1.6M daily transactions, earning a Sprint Star Award for analytics middleware that cut debugging time from 2 hours to 15 minutes.
Published ML researcher. Open-sourced methodology actively used by the marine biology research community. MSc Data Analytics, Queen Mary University of London.
Frontend was the beginning — it built product instincts and taught how users interact with data. Every system at Barclays and US Bank had a data architecture problem underneath the UI problem. The MSc formalised what was already happening: owning the full system, not just the interface layer. The role the market is starting to name is AI Product Engineer — one engineer who designs the data model, builds the pipeline, integrates the intelligence layer, and ships the interface. Seven years of deliberately building across all three layers is exactly the preparation for it.
End-to-end analytics platform built for Barclays Investment Banking — Python ELT pipelines on AWS Glue processing 150K–200K financial records per batch, star schema semantic data models, Tableau executive dashboards, and LLM-powered reporting interfaces replacing 72-hour manual workflows.
GraphQL Backend-for-Frontend with Apollo Server designed from scratch covering all treasury payment flows. Subscription-based WebSocket connections for concurrent OLTP state consistency. 40 React/TypeScript components, LLM-powered conversational features, A/B tested across 340K MAU.
60-component React/TypeScript design system built from scratch with design token architecture, Storybook documentation, and Chromatic visual regression testing. Published as npm package. Vite migration cut bundle by 39%, load time by 45%. Served 8,000+ daily users across 8 product teams.
MSc dissertation: domain-generalised image re-identification model eliminating identity leakage bias in computer vision. PyTorch, TorchReID, OSNet with Grad-CAM interpretability. Two papers published. Open-sourced evaluation methodology actively used by the marine biology research community.
Open to Senior Software Engineer roles in London — data engineering, full-stack, product engineering.
Eligible for UK Skilled Worker sponsorship. Available immediately. Currently targeting Bloomberg, JPMorgan, Snyk, Airbnb, and Attio — always open to the right conversation.