AI Call Centre Case Study
Whether you have a detailed brief or just a problem worth solving — this is the right place
to start.
The Challenge
A well-established East African media house had built a streaming platform that had served them in the
early days — but the cracks were showing. Content delivery costs were climbing as the subscriber base
grew, buffering complaints were frequent across mobile networks in lower-bandwidth regions, and the
editorial team had no way to publish or update content without involving developers. What had started as
a functional system had become a bottleneck.
The business needed a platform that could handle growth without proportionally growing infrastructure
costs — one that could deliver reliably on the patchy mobile connections that the majority of their
audience used, and give their content team genuine independence.
What We Built
CDN Architecture Overhaul
- Migrated content delivery from a single-origin model to a multi-CDN architecture using AWS CloudFront with intelligent edge caching
- Implemented adaptive bitrate streaming (HLS/DASH) - content now automatically adjusts quality to the viewer's available bandwidth rather than buffering or failing
- Configured regional CDN edge nodes optimised for East African network topology, reducing average latency to end-users by routing through geographically closer points of presence
Transcoding Pipeline
- Replaced a manual, developer-dependent transcoding process with an automated pipeline built on FFmpeg - an open-source transcoding engine customised for the platform's specific content formats and delivery requirements
- Content ingested in any format is automatically transcoded into multiple bitrate profiles - 240p through 1080p - ready for adaptive delivery
- Proprietary transcoding pipeline built in-house means no dependency on managed cloud transcoding services - full control over encoding parameters and cost structure
Editorial CMS & Content Ingestion
- Built a purpose-designed content management interface for non-technical editorial staff - upload, tag, schedule, and publish content without developer involvement
- Metadata management, content categorisation, and scheduling built into a single workflow with role-based access for editors, reviewers, and publishers
- DDEX integration — the CMS supports content ingestion using the DDEX standard (Digital Data Exchange), the global industry standard for digital media supply chain communication, enabling automated ingestion from labels, distributors, and content aggregators
Multi-Device Applications
- Rebuilt web frontend in Angular - performance-optimised, role-based access for editorial and management functions
- Mobile apps (iOS and Android) rebuilt in Flutter - single codebase delivering native performance across platforms, with offline caching for content browsing in low-connectivity conditions
Recommendation Engine
- Built a content recommendation engine using LensKit - an open-source recommendation toolkit - trained on subscriber viewing behaviour to surface personalised content suggestions
- Collaborative filtering and content-based models combined to improve content discovery and increase session depth
Outcome
- Content delivery cost reduced measurably through CDN architecture and in-house FFmpeg transcoding pipeline. Buffering complaints dropped significantly. Editorial team publishes content independently — DDEX-standard ingestion enables automated delivery from labels and distributors.
Technology Stack
Three continents. Five platform domains. Real clients who trusted us with business-critical
systems. Here is what that produced.
Streaming
- HLS/DASH adaptive bitrate
- FFmpeg (proprietary in-house transcoding pipeline)
CDN
- AWS CloudFront — multi-region edge configuration
Frontend (Web)
- Angular
Mobile
- Flutter - iOS and Android
Recommendation
- LensKit - open-source collaborative filtering engine
CMS & Ingestion
- Custom-built editorial platform with DDEX standard content ingestion
Backend
- Node.js (NestJS)
- Java Spring Boot
Database
- MySQL
- Redis
Building or Rebuilding a Streaming Platform?
Let’s talk about architecture, cost, and what a rebuild would look like for your situation.