Case Study

Building an Intelligence Layer Inside a Cloud-Native Media Production Platform

Written by Craftsmen Research Team | May 11, 2026 11:00:55 AM
15+
Countries with active Mimir deployments
600K+
Celebrities detectable via AI recognition
9+
Years of embedded engineering partnership
99.9%
Platform uptime for broadcast-grade workflows

A Norwegian media tech company building the future of newsroom production

Mimir is a cloud-native media production and collaboration platform built by Mimir Media Tech AS, part of Norway’s Fonn Group. It gives journalists, editors, and producers one cloud environment for production asset management, archive, AI-powered metadata enrichment, live production, and editorial collaboration.

Its clients include TV 2 Norway, VRT, Sky News, RTL, TVNZ, Canal 9, DPA Deutsche Presse-Agentur, the Philadelphia Eagles, and blinx, which reached 1 billion video views within 5 months of going live on Mimir.

This is not a MAM tool for small production houses. It is a production-grade cloud platform built for organisations where content must be found, edited, and published against live broadcast deadlines.

The Challenge

From a Capable Archive Tool to an Intelligent Production Platform Without Dropping a Frame 

When Mimir was still in its MVP phase, Craftsmen was already part of the engineering journey. Together, the teams were not simply building another cloud storage tool. They were laying the foundation for a production asset management platform that could evolve with a media industry moving rapidly from on-premise infrastructure to cloud-first workflows.

The challenge became multi-dimensional as Mimir scaled. The platform needed to orchestrate multiple AI services simultaneously and make them work seamlessly at ingest scale, handling libraries of 50,000+ hours of content without slowing down or introducing metadata errors. At the same time, it had to preserve the reliability broadcast clients depend on. When a journalist at Sky News searches for a clip of a specific person speaking about a specific topic in a specific location, the system has to find it fast, accurately, every time.

The platform had to support demanding broadcast workflows, scale across large media libraries, and maintain the speed, reliability, and accuracy expected by tier-one clients.

This required deep technical ownership across critical areas of the product, from cloud-native infrastructure and ingest workflows to AI-powered metadata, search performance, and editorial collaboration. The work was not simply about adding development capacity. It was about building and evolving a production-grade platform that broadcasters could rely on every day.

 The Development Partnership 

How Craftsmen Engineers Owned Mimir’s AI Pipelines Under Broadcast Pressure

 

AI-Native Engineering

How Craftsmen engineers worked at 2026 speed, not 2022 speed

Across the Mimir engagement, Craftsmen engineers used AI-assisted development tools as standard practice. AI-driven code generation accelerated boilerplate-heavy integrations across multiple AWS SDK interactions, while AI-assisted test generation improved coverage across metadata processing pipelines, especially for edge cases involving unexpected media formats or AI service response variations at scale.

Code review workflows were also augmented with AI analysis to catch performance anti-patterns in the event-driven architecture before production, where a slow query against a 50,000-hour archive can directly affect journalist productivity during a live news event.

 

"The team Craftsmen embedded inside our product understood what broadcast clients demand. Not just technically, but in terms of the accountability and ownership that comes with building for newsrooms under live deadline pressure. They took real engineering responsibility and delivered it."

— Engineering leadership, Mimir Media Tech AS · Norway

Technology Stack Partnership Outcomes
AWS Lambda + SQS (event-driven processing) Multi-language speech-to-text search across full archive
AWS Rekognition (face + object detection) 600K+ celebrity recognition at ingest
AWS Transcribe + Translate Object and face search across video timeline
AWS Elemental Live Adobe Premiere integration with proxy switching
Adobe Premiere Pro Extension API Live transcription for live production workflows
Node.js / TypeScript Scale from 50 mins to 50,000 hours without re-architecture
PostgreSQL + Elasticsearch Deployed for TV2, VRT, Sky News, DPA, NRK, blinx
AWS S3 + MediaConvert  
React (web application)  

 

Business Impact

From archive tool to intelligence layer and the commercial proof that followed

With AI metadata enrichment running at production scale inside Mimir's platform, the product's commercial positioning changed materially. Broadcasters were no longer evaluating a cloud-hosted archive with search. They were evaluating a searchable intelligence layer that could find any person, speaking any words, in any language, across decades of content, instantly.

TV 2 Norway selected Mimir for all editorial media workflows,  a whole-organisation commitment from one of Norway's most significant broadcasters. VRT, Belgium's public broadcaster, migrated to Mimir for cloud-native news operations. NRK, Norway's national broadcaster, chose Mimir as their new cloud-native DAM. The Philadelphia Eagles brought Mimir to the US sports market. And blinx, a Middle Eastern digital storytelling hub, reached 1 billion video views within 5 months of launching on the platform.