AI Engineering Lead in London

The opportunity Substrate is building a network of fully autonomous wet labs, cloud-based data production facilities for AI biology, integrated with foundation models to become the critical infrastructure layer for AI-driven biological discovery. Our first node opens in King’s Cross, London, with several integrated workcells and two scientific verticals online by mid-2027. Our customers range from foundation model labs to global pharma. We are hiring an AI engineering lead as the first engineering hire on the intelligence software product.

Substrate runs two AI products on top of operational data today, with more to come. You will write much of the first production code, help shape the architecture, and influence how the team that comes after you is built.

About Substrate: Substrate is spinning out of Automata, the UK lab automation company that has built the workcell platform our labs run on. Our four co-founders are Mostafa ElSayed (CEO and founder of Automata), Oli Hoy (formerly VP Customer Experience at Automata), Alexey Morgunov (AI Scientist co-founder, leading the intelligence software product), and a founding biology lead joining shortly. We are aiming to have ramped up to 32 people by the end of Q1 2027. We are funded in parallel by a combination of venture funding and government grants. We are not a cloud lab and we are not a CRO. We are an autonomous lab platform with closed-loop integration available as one operating mode for foundation model partners.

The role: You will sit alongside Alexey on the intelligence software product. You will write most of the early production code, help shape the architecture, and influence the engineering culture as the team grows. The product has two surfaces today and will grow over the next year. AI Scientist is an agent that ingests the scientific literature, identifies inconsistencies and high-value gaps in published data, and flags experiments where Substrate’s reserved R&D capacity could resolve open questions. AI Assays is a continuous-improvement product that uses run metadata to optimise assay protocols over time, reducing variable cost and increasing throughput. Both products run on top of an autonomous wet lab that is generating its own structured operational data from day one. You will work closely with Alexey on technical direction, with the founding software engineer on the boundary between operational and intelligence software, and with the founding biology team and vertical leads on the scientific content that AI Scientist consumes.

What you will do in your first twelve months:

• PHASE 1: SEP TO DEC 2026 - Land in the team. Help lock the architecture for AI Scientist and the harness layer underneath it. Ship the first end-to-end thin slice: literature ingestion, gap identification, and a working agentic loop that surfaces candidate experiments. Help set the engineering culture for the intelligence team: code review, evaluation, deployment, observability. Influence the languages, frameworks, and tooling we will live with.
• PHASE 2: JAN TO MAR 2027 - Stand up the data capture infrastructure for AI Assays. Define the schema and the feedback path back from operational software. Bring AI Scientist into production. Wire its outputs into how Substrate allocates the reserved R&D capacity across verticals. Help shape the technical roadmap for the intelligence team as the AI engineer and data engineer come online alongside you.
• PHASE 3: MAR TO JUN 2027 - Ship the first version of AI Assays. Connect protocol-optimisation suggestions back into the assay design loop. Help scope the third intelligence product on top of the data the lab is now producing at scale. Move from primarily writing code to helping coordinate the intelligence team’s work across product surfaces.

Who you are: You are an experienced software engineer who has put large language models, foundation models, and agentic systems into real production, not as a prototype or a demo. You know the harness layer well: token economics, retrieval, evaluation pipelines, structured output, the operational realities of running a lot of data through LLMs cheaply and reliably. You enjoy that work. You have some history with biology, biotech, or scientific literature. That can be a formal background, a previous role at a science-adjacent company, or simply that you have read papers in depth, kept up with the field outside of your day job, and have a feel for what experimental data telemetry actually looks like. You do not need a PhD; you do need to be the kind of engineer who finds the science genuinely interesting. You are direct. You will talk back when you disagree. You are pragmatic about agentic systems and foundation models; you have used them in anger rather than read about them in posts.

MUST HAVE:

• Five or more years of professional software engineering experience.
• Direct experience putting LLMs, foundation models, or agentic systems into production at scale.
• Working comfort with the LLM harness layer: token economics, retrieval, evaluation, structured output, large-scale data processing through models.
• Strong working comfort with Python.
• Track record of designing systems that other engineers built on top of.

NICE TO HAVE:

• Direct experience in or near biology, biotech, scientific computing, or a research environment where experimental data and academic literature were part of the day job.
• Experience of an early-stage founding-engineer role at a venture-backed company.
• Background near LIMS, ELN, or scientific data infrastructure systems.

Why this is unusual: Most AI engineering roles at venture-backed companies are either pure AI applications (chat products, copilots, agents on top of someone else’s data) or thin wrappers around foundation model APIs. This is neither. You will be building AI products on top of the operational data of a wet lab that you can sit next to and influence the design of. AI Scientist decides which experiments are worth running with Substrate’s reserved R&D capacity, by reading the scientific literature and identifying what has not been done well. AI Assays makes the lab better at its own work every week, from the operational metadata of every run. The closest analogue is the internal tooling team at a frontier model lab, with one important difference: you control the data source. Some engineers find this energising; some find it distracting. Worth knowing in advance which one you are.

Compensation and equity: We pay competitively against the London market for senior engineers working on LLMs and agentic systems at venture-backed companies, calibrated to seniority and to the specific scope of this role. We will discuss numbers with serious candidates after first conversations. Equity is meaningful, with vesting on the standard four-year schedule and a one-year cliff. We can talk through the philosophy and the maths in detail when we meet.

How we work: Working pattern is open. We will design around the strongest candidate, with a bias towards willingness to spend some in-person time at our King’s Cross site, particularly during the early phases while the team is forming and the architecture is being set. Most of the founding team are in the office most days. 30 days annual leave. A learning budget you can use for conferences, courses, books, and time. The founding team operates on a weekly cadence with a Monday planning meeting and a Friday close, and a quarterly offsite. We are direct with each other, we write things down, and we expect to be challenged.

The team you will join: You will report to Alexey Morgunov, co-founder, which leads our intelligence software product. You will work most closely with the founding software engineer on the boundary between operational software and intelligence software, and with the founding biology team and the protein and functional genomics vertical leads on the scientific content that AI Scientist consumes. You are the first intelligence team hire alongside the bio-AI specialist, with a junior AI engineer and a data engineer joining shortly after. Substrate is currently four co-founders growing to 32 people by Q1 2027.

How to apply: Apply via Ashby with whatever you think shows your work best: a CV, links to GitHub or to systems you have built, a piece of writing you are proud of, an evaluation harness you ran on a model that taught you something … We read everything that comes in. Our process is four stages. An initial conversation with Alexey to understand what you want from the role and what we want from it. Two technical sessions with our external technical advisor: an architecture deep-dive on how you would build the intelligence software, and a session on how you would build and grow the intelligence team. Finally, an in-person founder-team session covering scope, terms, and any final questions. We aim to move fast on candidates we are excited about; expect roughly two to three weeks end to end. If you are not sure whether you are a fit, send a note anyway. The most useful conversations we have had so far have been with people who were not sure. Substrate is an equal opportunity employer. We make hiring decisions on merit, scope-fit, and the strength of the working relationship we expect to build with each hire. Applications welcome from candidates of any background.