Rigorous software for unusual problems.
We design, build and run the software that research and engineering organisations depend on: data pipelines, scientific platforms, modelling tools and serverless cloud infrastructure on AWS.
We build the data pipelines your scientists are currently running in Excel.
Scientific software & data platforms
Pipelines, analysis portals and research tooling that scientists actually use.
Learn moreServerless & AWS architecture
Scale-to-zero, event-driven systems, infrastructure-as-code, cost and reliability reviews.
Learn moreFull-stack product engineering
Web applications from MVP to production in Python and TypeScript.
Learn moreModernisation
From spreadsheets, scripts and on-prem HPC to maintainable cloud-native software.
Learn moreA small, senior team that trained as scientists and engineers first.
Rigour over hype
We measure, test and write it down. If we can't explain why it works, it isn't finished.
Curious by default
Unusual problems are the interesting ones. We go and look, then dig until we understand the cause.
Candid
We say what broke and how we fixed it, to clients and in public. Honest trade-offs beat confident guesses.
Built to be owned
Documented, tested, infrastructure-as-code. You should be able to run what we build without us.
- Python
- Django
- FastAPI
- TypeScript
- Svelte
- Tailwind CSS
- PostgreSQL
- MongoDB
- Docker
- AWS
- AWS CDK
- Stripe
Working with Weird Sheep Labs has been an absolute pleasure. They quickly grasped our challenges and the intricacies of our existing product, seamlessly adapting to it. The structured approach and professionalism have made the collaboration smooth and highly effective. I couldn't recommend more to work with Weird Sheep Labs!
Working with Weird Sheep Labs was a smooth experience! They successfully built and deployed the MVP for our platform, working across the stack on frontend UI, backend APIs, devops and system architecture.
Weird Sheep Labs is an outstanding consultancy partner who has consistently exceeded expectations. Their contributions have extended far beyond the initial project scope, adding significant value across multiple areas of our organization. Their performance has been exemplary, and I would highly recommend their services without reservation.
From the blog
What we've learned building serverless systems and scientific software, including what went wrong.

How a one-line CDK Aspect took down every Lambda in our dev environment
At Weird Sheep Labs, we lean heavily on containerised Lambda functions. On a recent client project, a seemingly harmless CDK Aspect (all it did was turn on image scanning) took down every Lambda in the backend dev environment overnight. The culprit turned out to be a CloudFormation behaviour that anyone running Docker-image Lambdas on CDK can trip over, and one that neither deploy output nor drift detection will warn you about. Here's what happened and how to avoid it.

Handling Lambda invocation timeouts in batch processing
Building serverless data pipelines with AWS Lambda, SQS, and DynamoDB has become a common pattern for processing large volumes of messages. The architecture is straightforward: SQS delivers batches of messages to Lambda, which processes each one and tracks its status in DynamoDB. When something goes wrong with a message, Lambda marks it as failed in the database, allowing it to be retried. However, Lambda timeouts can break this pattern in a critical way, orphaning SQS messages and leaving your system in an inconsistent state.

Using FFmpeg in AWS Lambda with Docker
Audio and video processing in serverless environments like AWS Lambda has traditionally been challenging due to the lack of native support for media tools. However, we can now package powerful media processing libraries like FFmpeg directly into our deployments by containerising Lambda functions with Docker. In this post, we'll explore how to effectively use FFmpeg within AWS Lambda by examining a proof of concept transcription service we recently built for a client.



