Case Studies | SaaS B2B

Live meeting transcripts that highlight what each person cares about

LLM Live Transcription PoC

About

A SaaS B2B business working with Cloud Combinator on AWS. The client is anonymised at their request.

Challenge

The proof of concept had three focus areas, each tied directly to the project's success criteria.

Controlling the live recording

The service needed a clean way to start and stop transcription of a live meeting on demand, with each session tracked from the moment it went active to the moment it stopped. Without reliable session control, nothing downstream could be trusted.

Per-user highlighting in near real time

A large language model had to classify chunks of the running transcript against each viewer's stated interests, and do so within a tight time frame while the stream was still live. At meeting pace, a highlight that arrives late is a highlight that has already been missed.

Reliable logging and retrieval

Every transcription session, user query and processed highlight needed a sound logging strategy so viewers could retrieve their real-time highlights on demand and the team could reason about what the pipeline had done.

Solution

An administrator starts or stops a session through Amazon API Gateway, which invokes an AWS Lambda function that opens or closes the stream and starts or stops the job in Amazon Transcribe. Viewers submit the topics they want highlighted through a second API, and those preferences are stored against the session in Amazon DynamoDB.

10s

Target processing cycle for real-time highlights

4

Lambda functions across the end-to-end pipeline

Claude 3

Bedrock model chosen to power per-user highlighting

By the numbers:

  • 10s - Target processing cycle for real-time highlights
  • 4 - Lambda functions across the end-to-end pipeline
  • Claude 3 - Bedrock model chosen to power per-user highlighting
Changes

The proof of concept met its acceptance criteria: the pipeline could start and stop live transcription, capture per-user preferences, process the running transcript on a timed cycle, and use Amazon Bedrock to extract and log per-user highlights, all deployed in a secure Amazon VPC and delivered as infrastructure as code. It confirmed AWS as a suitable environment for the client to continue developing the service.

  • Live transcription proven — Sessions could be started and stopped on demand through an API, with Amazon Transcribe producing the running transcript for each active stream.
  • Personalised highlights — A model in Amazon Bedrock classified transcript chunks against each viewer's stated interests, returning highlights tailored to the individual rather than a single generic view.
  • Event-driven and serverless — An EventBridge rule and a set of Lambda functions kept processing on a short, repeating cycle, with DynamoDB providing low-latency storage and a full log of the pipeline's activity.
  • Secure, reproducible foundation — The stack was deployed with AWS CloudFormation into a secure VPC with backups configured, giving the client a clean base rather than a throwaway prototype.
  • Production path defined — The engagement set out clear recommendations for scaling, security, observability and cost governance to take the service from proof of concept toward production.

With the pipeline validated and a production roadmap in hand, the client has a proven AWS foundation on which to harden the service, tune highlight accuracy and scale toward live audiences.

AWS Stack

Amazon Transcribe

For live, streaming speech-to-text of the meeting audio.

Amazon Bedrock

For the large language model that classifies transcript chunks and selects per-user highlights.

AWS Lambda

For the serverless functions that control sessions and orchestrate the real-time processing.

Amazon DynamoDB

For low-latency storage of transcripts, user preferences and highlights, with backups.

Amazon API Gateway

For the endpoints that start and stop streams, capture preferences and return highlights.

Amazon EventBridge

For triggering the processing cycle on active transcription sessions.

YOU MIGHT LIKE

Related success stories

View all case studies

Case Studies | Insights

Utilising Language Recognition, Speed, and Enhanced Security to Make Social Media a Force for Good

  • Here, we take a detailed look at how the Cloud Combinator team collaborated with another cutting-edge AI service provider that provides intelligent systems to “make social media more social” for brands and users alike.
  • Arwen AI is a UK-based startup specialising in AI solutions to manage and enhance brands’ social media interactions. Founded in 2020 by Matt McGrory, Dr. David Cole, and Joel Bailey, Arwen. AI focuses on using AI to automatically detect and remove spam, toxic comments, and other unwanted content from social media platforms.
  • The team at Arwen have three core products. ‘Moderate’ is focused on identifying and removing toxic content from social media channels. ‘Engage’ helps brands identify and engage with meaningful conversations on social media, and ‘Customize’ allows brands to apply bespoke algorithms to their channels - creating an even more effective moderation and engagement.
Read more
CONTACT US

Ready to turn AI into impact?

We'll help you spot the highest-value opportunities, reduce risk around your first AI initiative, and define a clear path to results from day one.

Why talk to us:

Outcome-driven recommendations

AWS-recognised delivery expertise

Risk-aware AI adoption

Clear next step, not a sales pitch

Start with a focused 20-minute conversation about your goals — no pressure, no commitment.

We use cookies

This website uses cookies to enhance user experience and to analyze performance and traffic on our website.

See our Privacy Policy for details.