Case Study: Global telecommunications provider

Designing and building an automated speech recognition pipeline

Topics

Kubernetes Serverless Optimisation

Industry

Media Enterprise

Key Technologies

Challenges

Our client provides video subtitling services to a number of broadcasters in multiple languages, with over 100,000 hours a year of live captioning. Subtitle generation is often a manual and slow process as well as an expensive one and LiveWyer helped to reduce their running costs.

Solutions

The use of AI driven speech-to-text systems has the potential to disrupt the market for traditional subtitling services. The company engaged with LiveWyer to build a container based solution for an Automatic Speech Recognition pipeline and batch processing platform capable of automating the generation of Spanish and English subtitles.

Results

LiveWyer was able to take a core business idea and evolve it into a series of key technology assets with cost, scale and outcome benefits to the client.

0

Idle running costs

60%

License cost reduction

3

months total engagement

Full Report

Discovery

It was our task to take the client’s high-level end state and develop a series of features, stories, outcomes and feed this into a development roadmap. By engaging in a series of discussions and demos, whilst utilising Serverless technologies, we were able to rapidly deliver a functional proof-of-concept platform at the end of this extended discovery phase. This technology demo then helped direct further discussions as we moved into the main development phase.

Development

With the core concepts proven, we began development of our production build, which needed to fulfil a number of business goals in terms of scalability, cost reduction and regulatory compliance. This required us to extend the Serverless platform and integrate with an existing on-premise Kubernetes platform we had previously delivered. This meant that workloads could be managed and secured within their existing infrastructure governance.

Cost Reduction

As well as additional scaling benefits, the client’s move to Kubernetes allowed them to benefit from a cost-effective licensing arrangement provided by a key third-party service. By moving from a per-request to flat per-core model with Kubernetes, the client benefited from a substantially reduced per-conversion cost whilst still being able to manage the service at scale within their current operational models. Another benefit was that by using an existing Kubernetes control-plane, they were able to keep their idle hardware costs at zero whilst only scaling resources when existing capacity was exhausted and further jobs demanded it.

Results

LiveWyer was able to take a core business idea and evolve it into a series of key technology assets with cost, scale and outcome benefits to the client. By aligning the right technology with the right requirement, LiveWyer were able to successfully implement and extend Serverless technologies as well as migrating workloads to compliant infrastructure. We were able to demonstrate how we can adapt our solution to meet regulatory and financial requirements.

Click here to download a PDF summary of this case study