2 days ago
London, United KingdomSenior
Responsibilities
- Develop hardware-aware transformations that adapt models to different accelerator architectures and execution environments.
- Explore numerical, structural, and memory-related optimisations for model variants with different capability, latency, and cost trade-offs.
- Investigate inference algorithms, including speculative decoding, for heterogeneous computational resources.
- Develop automated methods to discover, characterise, and optimise hardware-specialised model variants.
- Build rigorous evaluation methods to measure model quality, efficiency, and the effects of architectural modifications.
Requirements
- Deep understanding of modern LLM architectures, including attention mechanisms, model structure, and inference implications.
- Experience in model architecture research, efficient inference, or hardware-aware machine learning.
- Familiarity with architectures such as mixture-of-experts, diffusion, JEPA, or alternative attention mechanisms.
- Strong intuition for interactions between model capability, numerical behaviour, memory requirements, and execution efficiency.
- Experience designing rigorous experiments to assess model quality, efficiency, and architectural modifications.
Benefits
- Competitive salary determined by skills and experience.
- Equity and ownership.
- Private healthcare.
- Visa sponsorship and relocation benefits.
- In-person work at the company’s London office with provided tools, space, and workspace setup.
Categories
AI ResearchML Engineering
About Callosum
Callosum builds AI infrastructure that unifies heterogeneous compute—software and systems to orchestrate diverse accelerators and the physical clusters they run on—for organizations training and serving models. The privately held company was founded in 2025 and is headquartered in London. It focuses on designing, deploying, and operating mixed-hardware compute fleets across sites and vendors, and raised a seed round in 2026.