Stand out for this role — generate a tailored resume and cover letter in about a minute.
CommonAI is seeking a capable software engineer to join the Compute team in Cambridge. You will build and ship features for multi-cloud AI compute products, using agent-assisted development to accelerate delivery while maintaining code quality and security.
The role suits an early-career developer who can reason about systems, learn quickly, and work across Linux, containers, and cloud environments with senior review on higher-risk changes.
CommonAI is building shared infrastructure and products that help organisations develop and deploy deep-tech AI systems.
CommonAI Compute, part of the CommonAI ecosystem, develops multi-cloud compute products and services for AI teams. We work with commercial, government and research partners to accelerate the adoption of AI across the UK and Europe.
We are a member of the UK government-funded £70 million Scaling Inference Programme, which is advancing AI inference hardware and software infrastructure. CommonAI is also backed by Barclays through programmes focused on deploying AI in regulated markets. Work across these programmes informs and supports our commercial product development.
We are looking for a capable software engineer who combines strong engineering fundamentals with an enthusiastic, practical approach to AI-assisted development. You should be an exceptional early-career developer who learns quickly and has a strong record of technically demanding builds.
This is not a role for someone who uses AI without understanding the code or is limited to occasional autocomplete. We want engineers who can reason about systems independently, then use agents to explore, implement, test and ship substantially faster.
Agent-assisted engineering is core to the role, not optional. We want engineers who can break ambiguous problems into agent-executable work, give agents the right context and constraints, challenge and verify generated code, and know when to take back control. What you build matters, not how many prompts it took.