Making AI the everyday way of working for an engineering team
Connected-TV analytics
Claude Code became the team's everyday way of working, with changes still going through normal pull-request review.
A cloud platform that matches TV content at scale
Connected-TV analytics
Key figures
- 200+ capture devices
- 6 million+ titles in the catalogue matched against
- 17.7 million artworks in the catalogue matched against
The challenge
The business needed to capture what appears on screen across a large set of consumer devices and match it, reliably and at scale, against a very large catalogue of titles and artwork.
What we did
- Architected and built an AWS platform using Lambda, AWS Batch, SQS fan-out and Step Functions, defined as infrastructure as code in CloudFormation and deployed through GitHub Actions.
- Built a GraphQL API with more than 80 resolvers over PostgreSQL, with Cognito sign-in, plus a machine-to-machine REST data API.
- Matched content using perceptual hashing and vector-embedding similarity search, with confidence-scored fallbacks, and rules engines that assign and publish results automatically.
- Added a ClickHouse analytical mirror and Grafana dashboards kept as code.
The result
The platform matches content from more than 200 capture devices against over 6 million titles and 17.7 million artworks, handling about 300,000 observations a week.
More case studies
Safer, faster change for a national public Wi-Fi platform
UK public Wi-Fi provider
Releases reached a daily cadence, on a platform that ran AAA accounting for millions of concurrent users.
Let's talk about what AI can do for your team
Tell us where you are today. We'll come back within one working day with practical next steps.
Start a conversation or email info@fife-v.com