ML Engineer at Artificial Societies in London

About Us

Artificial Societies (societies.ai) helps Fortune 500 organisations understand how real-world audiences think, feel, and respond — without the cost, delay, or limitations of traditional research. We build large-scale simulated populations grounded in real-world data, enabling clients to test messages, strategies, and concepts. Our clients love the ability to receive insights backed by millions of responses in hours rather than months, and to access previously inaccessible audiences (e.g. investors, CEOs, opinion leaders).

We're backed by Y Combinator, Point72 Ventures, and Kindred Capital, plus investors from DeepMind and Sequoia Scout. We work with leading organisations across technology, financial services, communications, and public affairs.

We're a lean and highly effective team. We move fast, hold a high bar, and trust each other to own outcomes end-to-end. The culture is collaborative but low-ego: ideas win on merit, not seniority.

The Problem

We simulate societies: how thousands of specific individuals influence each other and form opinions, across a multiverse of scenarios.

The combinatorics are as hard as protein folding. Give 1,000 people a for-or-against position on 10 issues and there are 103,000 possible group opinion states. DeepMind only has to model 10300 ways a protein can fold. Searching that space usefully — and fast enough that a client gets an answer in hours, not months — is a frontier machine learning problem.

To solve this, we are building a Societal World Model — an AlphaFold for societies. We are already informing strategies for clients with $3T+ in combined market caps. For a few seconds during one client simulation, we were 2% of Google's global AI throughput.

The Role

We're looking for an ML Engineer to own the path from frontier research to production. Our research team invents novel formulations of how opinion, influence, and belief move through a population; you turn them into production-grade systems that deliver that research to the world's most important organisations.

You'll take new simulation methods from prototype to production: post-training and distilling models that power them, making them fast and efficient to serve at scale, building the data pipelines that feed them, and shipping the whole system. You'll work directly with our CEO and Head of Research, and partner closely with our research, product, and forward-deployed engineers.

This is engineering that runs on ML judgement. A new method can arrive with six ideas in it; you'll know which one performs best, and you'll build the most simple and reliable version of it. You'll be in charge of deploying the world's most advanced social simulation research into production, where it will shape decisions that shape the world.

What You'll Do

Take research to production. Own the path from a working prototype to a deployed capability our clients depend on.

Post-train and distil models. Fine-tuning, RL, and distillation to make our simulations nuanced and accurate.

Own inference at scale. Serve simulations of thousands of individuals efficiently and reliably.
Build the data engineering. Design the pipelines that turn messy real-world data into simulated populations at scale.

Simplify aggressively. Reason about which parts of a complex method actually deliver the performance, and delete everything else.

What We're Looking For

At least 2+ years shipping ML in a startup environment, ideally as an AI or ML engineer at an early-stage company where you owned things end-to-end. You've put models in front of real users, not just in a notebook.

Deep learning, not classical ML. Hands-on with pre-training or post-training language models — RL, fine-tuning, distillation, dataset and reward design.

Strong Python engineering experience. You can write clean and well-structured Python code that other people can maintain.

Comfortable with hard concepts. You reason fluently about statistics, probability, and have intuitions around high-dimensional spaces.

Fast, with a high bar. You can ship quickly, and you go deep to understand the problem and the solution.

Deeply interested in what we're doing. Simulating how people form opinions is a social science problem as much as an ML one, and the best people here find that fascinating rather than incidental.

Ownership mentality. You don't wait for a ticket. You see what needs to happen, and you make it happen end-to-end.

Nice to Have

You've distilled a frontier model into a small open-weights one and shipped it to production to cut cost or latency.

Experience with distributed training, inference optimisation, or scaling ML systems.

Background in behavioural science, computational social science, psychometrics, statistics, or a related empirical field.

Publications or open-source work in LLM simulation, evaluation, calibration, or human-behaviour modelling.

Familiarity with TypeScript (our product stack) to collaborate smoothly with product engineers.

Logistics

Location: London, in-office. We build better together, and this role is no exception.

Salary: £100,000 – £180,000 depending on experience.

Ignore stated salary range with job posting - wellfound does not natively allow our true range which is £100,00-£180,000

Equity: Up to 0.20%.

Why Join Now

If you want to work on something truly frontier, this is it. You'd be working directly with some of the smartest people on the planet, on one of the hardest problems humanity has yet to solve. This is the role that carries frontier research into robust engineering capability — the work that delivers our inventions to organisations whose strategies shape the world.