Research Engineer, Electric Twin
£70-125k
4-5 days a week in office
Simulating human behaviour with large language models
Posted today
Simulating human behaviour with large language models
1-20 employees
Posted today
£70-125k
4-5 days a week in office
1-20 employees
Company mission
To improve our understanding of the world.
Role
Who you are
- We are looking for a creative and rigorous engineer who is passionate about using AI as a tool to understand the human world
- MS or PhD in Computer Science, Machine Learning, Computational Social Science, or a related quantitative field, or equivalent practical experience
- Deep experience in applying large language models (LLMs) to solve complex, open-ended problems
- Strong proficiency in Python and common ML libraries/frameworks (e.g., PyTorch, JAX, TensorFlow, Hugging Face)
- A strong scientific mindset geared towards rigorous experimentation, coupled with the engineering discipline to build the reliable and scalable systems needed to test complex hypotheses
- A keen interest and/or background in the principles of human behaviour, cognitive science, psychology, or social dynamics
- Excellent communication skills, with the ability to articulate complex technical and experimental concepts clearly
Desirable
- Experience or strong interest in multi-agent systems, agent-based modeling (ABM), or game theory
- Hands-on experience with advanced LLM application techniques like RAG, chain-of-thought, and agentic tool use
- Experience designing and conducting experiments in a social science or human-computer interaction (HCI) context
- A track record of publications in relevant AI or interdisciplinary conferences
- Experience deploying ML-driven applications into production environments
What the job involves
- As a Research Engineer at Electric Twin, you will focus on one of the most exciting and challenging applications of AI: modeling human behaviour. Your primary role is to pioneer the application of existing LLMs to create believable, consistent, and scientifically-grounded AI agents.
- You will design the cognitive architecture of our synthetic agents, figure out how to evaluate their behaviour in ambiguous scenarios, and develop the experimental methods to validate our simulations. You will be at the forefront of an emerging field, tackling questions like: How can we use an LLM to give an agent a persistent memory? How do we ensure a population of agents behaves in a way that is statistically representative of a real demographic? How do we measure and validate "realism" when there is no simple ground truth?
- This is a deeply interdisciplinary role that combines creative AI application, rigorous experimental design, and insights from computational social science to bring our synthetic populations to life.
- Design AI Agents: Develop sophisticated prompting strategies, retrieval-augmented generation (RAG) systems, and tool-use frameworks to imbue AI agents with consistent personas, memories, and reasoning capabilities
- Develop Experimental Methods: Design and execute experiments to test and validate the behavioural outputs of LLM-powered agents against real-world data and established social science principles
- Create Novel Evaluation Frameworks: Build and implement robust methods for measuring the quality and realism of our simulations, especially on ambiguous tasks where traditional accuracy metrics don't apply
- Build the Simulation Platform: Contribute to the core architecture that allows us to run, observe, and analyse complex simulations with thousands of interacting AI agents
- Translate Needs into Research: Work closely with product and commercial teams to understand client challenges and translate them into focused research questions and technical solutions
- Stay at the Forefront: Research and apply the latest techniques in agentic AI, multi-agent systems, and LLM evaluation to continuously advance our platform's capabilities
Salary benchmarks
Our take
Electric Twin is a company utilizing large language models to simulate human behaviour. Through its LLM models, users can create synthetic populations that represent their target markets or research demographics in order to investigate outcomes in near real-time and make more informed decisions.
The company supplies a range of investigative tools; from qualitative polling and quantitative analysis, to simulated focus groups or even one-on-one discussions. It also enables ideas to be continuously refined through strong feedback, ensuring that strategy can always be backed by the latest data.
With such a unique proposition, Electric Twin stands out as one to watch in the rapidly developing generative AI sector. It is currently working to build out its team and, if its models are able to deliver the value suggested, there will surely be no shortage of success in its future.
Kirsty
Company Specialist at Welcome to the Jungle
Company
Company HQ
Marylebone, London, UK
Leadership
Alex Cooper
(Co-founder)Former COO at UNDO, Director of Mass Testing for the Department of Health and Social Care, and Advisor for 10 Downing Street. Was also an Officer in the British Army for 19 years.
Ben Warner
(Co-founder)Former Chief advisor on Digital and Data to the Prime Minster with the UK Civil Service. Worked in Data Science at Faculty and was an EPSRC Research Fellow at UCL.
Share this job