LLM-Guided Evolutionary Optimisation Engineer

Callosum; Greater London; On-site; GBP 90,000 - 130,000; Full time

Benefits offered by this job

Equity & Ownership

Private healthcare

Visa sponsorship

London office

Job summary

Callosum is building intelligent systems that co-evolve hardware and software to run across heterogeneous workloads and accelerators. We are crafting an LLM-guided search loop that iteratively generates, evaluates, and improves programs against verifiable objectives.

You will develop the loop as a reusable internal optimisation system and a research object, improving search, measurement, and capabilities on increasingly difficult problems, kernels, runtimes, and scheduling.

Qualifications

• Strong Foundations in optimisation and search methods.
• Ability to reason about exploration vs exploitation and diversity.
• Experience turning experimental systems into reliable software.
• interdisciplinary approach across optimisation, algorithms and ML, not relying solely on LLMs.

Responsibilities

• Develop the core program evolution loop, candidates generation, evaluation, and evolution over long optimisation runs.
• Design search strategies balancing exploration, exploitation, diversity, and sample efficiency across large spaces.
• Build continuous benchmarks to measure the optimiser across internal workloads and standalone problems.
• Turn the system into reliable internal infrastructure that can scale to new problems.

Skills

Optimization basics

Exploration vs exploitation

Engineering instincts

Non-gradient methods

Job description

Callosum is building intelligent systems that co-evolve hardware and software to run across heterogeneous workloads and accelerators. We are crafting an LLM-guided search loop that iteratively generates, evaluates, and improves programs against verifiable objectives.

You will develop the loop as a reusable internal optimisation system and a research object, improving search, measurement, and capabilities on increasingly difficult problems, kernels, runtimes, and scheduling.