📍 Location: London, UK | San Francisco, CA
Canopy is building real-time voice and avatar models that bring computers to life.
About the role
You will design and operate the petabyte scale data infrastructure that drives large-scale training for our models. This role carries end-to-end responsibility for how data is ingested, filtered, structured, and delivered into training pipelines. Individual runs can represent millions of dollars in compute to just ingest and filter data.
This is a high-autonomy, high-trust role with direct influence on the foundation of our research stack.
What you'll do
- Architect and operate distributed pipelines that turn raw multimodal data into training-ready datasets at petabyte scale
- Build and operate distributed data pipelines with Spark/Ray
- Work with ML infra on performance-critical parts of the processing stack: model-assisted filtering, embedding generation, and evaluation signals
- Design storage + dataset layouts across S3 and GCS: formats, sharding, caching, IO patterns, etc.
- Operate production infrastructure with Docker
- Partner with research + engineering to translate evolving training requirements into robust, maintainable systems
You might be a fit if
- You have owned large-scale data infrastructure in production and care deeply about reliability
- You have strong opinions on databases and schema design
- You take pride in building simple, elegant and maintainable systems that execute with the highest standards