
Case studies
sub-200ms
500k+ docs indexed, 35% lower cloud cost
Rivian: RAG Knowledge Engine
An internal search system indexing 500k+ technical documents using hybrid dense + sparse retrieval, deployed on Kubernetes with sub-200ms latency.
Full funnel
Upload, customise, checkout, owned end to end
Custom Print Platform
A canvas-based design tool where users upload artwork, customise placement live, and order print products, built on Konva.js with WebSockets throughout.
4.1s → 1.8s
LCP across 2M+ monthly pages, conversion 2% → 4%
Marks & Spencer: Performance & Conversion
Rebuilt the rendering strategy and governed it with a 50–60 component design system across 2M+ monthly product pages.
Commerce, cloud platforms, and design systems.
99.9%
Uptime, MTTR down 30%
Enterprise banking platform
Cloud infrastructure and SRE for a regulated banking platform: 50+ production workloads on Kubernetes and OpenShift, automated GitLab delivery, and full observability across the estate.
2% → 4%
Checkout conversion, 40k+ users
Checkout Conversion Flow
Fixed fragmented checkout state and instrumented the full funnel with Mixpanel, then iterated on real drop-off data behind feature flags.
Excel-like
Frozen rows and columns
Siemens Teamcenter: Data grid
A spreadsheet-grade PLM data grid with frozen panes and rich inline editing, rendered with D3.js over an Apollo GraphQL backend.
2 brands
One themeable system
Cigora & Pipes and Cigars
A themeable design system on design tokens and atomic design, powering two distinct e-commerce brands from one source of truth.