Case studies

Real problems. Working systems.

These projects are described without invented customer names, revenue numbers or ROI percentages. Where a metric is not public, we omit it.

Computer Vision / Exhibition Analytics

Problem: A leading exhibition organisation needed to understand how people moved through a large venue — not just that cameras were recording. Playback was not the same as operational intelligence.
What we built: A computer vision system over existing camera infrastructure, covering people counting, movement analysis, dwell time and heatmap analytics across 100+ camera feeds.
Technology: Computer vision · People counting · Dwell time · Movement paths · Heatmaps · Analytics reporting
Business outcome: Stakeholders could measure attendance patterns, movement and exhibit-area activity from the cameras they already had — without replacing the CCTV infrastructure.

Healthcare / Clinical Documents

Problem: A healthcare technology company needed to process high volumes of clinical and operational documents that arrived unstructured, slowing downstream work.
What we built: Document processing pipelines for ingestion, structured information extraction and operational use of the resulting data — designed to sit inside existing healthcare workflows.
Technology: Document ingestion · Structured extraction · Validation workflows · Operational integration
Business outcome: Unstructured documents became structured information that operations could review and use. Exact throughput figures are held privately with the customer.

Insurance / Document Intelligence

Problem: An enterprise insurance workflow depended on PDF-heavy intake. Teams were reading, extracting and checking fields by hand before anything else could happen.
What we built: A document intelligence platform combining PDF ingestion, structured extraction, data pipelines, analytics dashboards and an end-user web application for review.
Technology: PDF ingestion · Field extraction · Data pipelines · Analytics dashboards · Review application
Business outcome: Insurance documents moved from unstructured files into a reviewable operational dataset. We do not publish customer-specific processing volumes here.

Enterprise Knowledge

Problem: Internal policies, SOPs, manuals and reports existed — but employees could not reliably find or trust an answer without asking a person who already knew.
What we built: An OCR ingestion pipeline, semantic search and a private knowledge interface so employees can ask questions of organisational documents and see the sources behind the answer.
Technology: OCR ingestion · Semantic search · Cited answers · Internal knowledge interface
Business outcome: Internal information became searchable in natural language, with source documents attached. The system was built for private organisational use, not a public chatbot.

Have a similar operational problem?

We will tell you whether we have built something close — and what a practical next step would be.