AI Engineer - AI Drug Discovery Company

Agentic AI Engineer — AI Drug Discovery Company (UK, Rare Disease Drug Discovery)

Up to £90k + Equity

Remote (Occasional visits to Cambridge Office)

We are working with an AI company applying artificial intelligence to drug discovery, with a focus on rare diseases that currently have no approved treatment. The company combines proprietary data, AI and deep scientific expertise to identify promising treatments faster and more cheaply than traditional drug discovery approaches.

This is not a green-field build. There is an existing agentic framework and set of workflows already delivering impact across the platform, and you will work alongside the engineer who owns that infrastructure, with scope to take ownership of specific workflows and grow your influence over time.

Key responsibilities include

Building and refining agentic workflows that help scientists generate the best testable hypotheses to progress.

Translating discovery problems into agent designs, and being clear-eyed about where an agentic approach adds value and where it does not.

Integrating agents with internal knowledge sources so they reason over the right evidence for each problem.

Expanding and maintaining existing GenAI tooling as the team's needs and the wider ecosystem evolve.

What we are looking for

You have shipped LLM-agentic systems that deliver real value in production, not just prototypes, with tools, multi-step orchestration and reliable behaviour.

You have at least two years of software engineering or ML engineering experience, with strong fundamentals and clean, tested, maintainable Python.

You are fluent in the modern LLM and agent toolkit, including model APIs, prompting, tool use, RAG, MCP, agent frameworks and evals.

You enjoy working closely with non-engineers and are comfortable sitting with a scientist to understand what they actually need.

Experience in drug discovery, biology or another life science domain is a bonus , as is familiarity with knowledge graphs, experience building evaluation harnesses for LLM systems, or a track record of picking up unfamiliar domains quickly.