Environmental policy

Efficient by default

We size models to the job rather than reaching for the largest option every time. Batch inference, caching and right-sized infrastructure cut the compute — and the energy — behind each campaign we run.

Cleaner infrastructure

Our workloads run in data-centre regions that report high renewable usage, and we schedule heavy training for lower-carbon hours where our providers expose that data.

• Renewable-backed hosting regions wherever available.
• Automatic shutdown of idle training instances.
• Annual review of provider sustainability reporting.

Accounting for the rest

We estimate the emissions we cannot yet remove and support verified removal and reforestation projects to cover them. We would rather cut first and offset second, and we publish both figures internally each year.

Talk to us

Working with us and want the footprint numbers for your account? Email hello@asoc.co and we will pull them together.