ML Research Scientist: Multi-Agent Environments & Harnesses

Lila Sciences; Greater London; On-site; GBP 170,000 - 227,000; Full time

Benefits offered by this job

Equity

Job summary

Lila Sciences is building a platform where AI and automation co-evolve to accelerate scientific discovery. We hire ML researchers to design synthetic RL environments and multi-agent harnesses that scale training and testing.

The role centers on creating novel harnesses, environments, and end-to-end pipelines for scalable experimentation. Ideal candidates have a PhD or equivalent, strong Python skills, and a proven ability to prototype and iterate in research settings.

Qualifications

• PhD in Computer Science or related field plus two years industry experience.
• Strong track record ideating harness designs or creating reinforcement learning environments at scale.
• Strong prototyping and coding skills in Python.
• Creativity and research intuition for identifying promising directions and designing experiments.
• Familiarity with core concepts and algorithms in reinforcement learning.
• Solid software engineering fundamentals: clean code, version control, testing, and integrating experimental work into shared infrastructure.

Responsibilities

• Design and prototype new harness capabilities — context compaction, novel tool interfaces, long-horizon planning, multi-agent swarms.
• Invent new RL environments by turning scientific workflows, simulators, and real-world tasks into training signals.
• Build end-to-end pipelines for synthetic environment generation at scale.
• Prototype and validate environment designs from concept through reward shaping and data collection.
• Advance multi-step reasoning with chained tasks and hierarchical goals.

Skills

PhD in CS

Reinforcement learning

Python

Software engineering

Education

PhD or equivalent experience

Tools

Ray

Modal

vLLM

APIs

Job description

Lila Sciences is building a platform where AI and automation co-evolve to accelerate scientific discovery. We hire ML researchers to design synthetic RL environments and multi-agent harnesses that scale training and testing.

The role centers on creating novel harnesses, environments, and end-to-end pipelines for scalable experimentation. Ideal candidates have a PhD or equivalent, strong Python skills, and a proven ability to prototype and iterate in research settings.