Case Studies | SaaS B2B

From Google Colab to a production ready inference API

Sound Recognition ML Platform

About

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

Challenge

The engagement had three focus areas.

Hosting a large, growing dataset

The training dataset was already around 50 GB, with room needed for future growth and augmentation. Google Drive could not support training at that scale, so the data needed a storage solution that was cost efficient and ready to grow.

Training the model cost efficiently

The neural network, written in PyTorch, needed to be rebuilt, retrained and evaluated on AWS in the most cost efficient way, with the flexibility to change or

Deploying a scalable inference API

The end product was an API that accepts a zip of one or more WAV files, runs each through the network and returns a JSON response of the most likely sound sources. It needed to scale cost efficiently from under a thousand calls a month towards several thousand, handle many files at once, and support per customer access keys, usage counting and short lived storage of uploaded audio for quality control.

Solution

We delivered the build as a clear path from data, through training, to a deployed API, implementing each stage with the client's engineer so the team could operate and extend it independently.

~50 GB

Audio dataset hosted on S3 with room to grow

1,000

Concurrent WAV files the solution is designed to handle

JSON API

Sound source predictions returned to manufacturers

By the numbers:

  • ~50 GB - Audio dataset hosted on S3 with room to grow
  • 1,000 - Concurrent WAV files the solution is designed to handle
  • JSON API - Sound source predictions returned to manufacturers
Changes

The client moved from a Colab based proof of concept to a scalable AWS architecture spanning dataset storage, model training and a deployed inference API, with the cost controls and know how to run it within their AWS credits.

  • Scalable data foundation — The training dataset moved to Amazon S3, removing the size limits of Google Drive and ready to accommodate an augmented, growing dataset.
  • Flexible, cost aware training — Model training and evaluation on Amazon SageMaker, with the freedom to resize the network and clear visibility of compute spend against AWS credits.
  • Production style API — An endpoint that ingests WAV files and returns JSON predictions, with per customer keys, usage counting and short lived upload storage for quality control.
  • Cost control and self sufficiency — The client were shown how to track and predict costs, scale compute up and down, shut resources down to avoid waste, and set spend warnings against their credits.
  • Handover — The architecture was implemented alongside the client's engineer, leaving the team able to change the model and operate the platform themselves.

With a scalable training and inference platform in place, the client are positioned to grow their model and dataset and to onboard noise monitor manufacturers to the API as demand builds, on an AWS foundation that scales cost efficiently with them.

AWS Stack

Amazon S3

For cost efficient, scalable storage of the audio training dataset.

Amazon SageMaker

For building, training and evaluating the sound recognition neural network.

Amazon API Gateway

For a secure, key controlled API for noise monitor manufacturers.

AWS Lambda

For serverless orchestration of audio unpacking and model inference.

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.