Locai Labs deploys AI coding assistant at First Light Fusion

Locai Labs has signed a contract with First Light Fusion to supply and deploy a domain-specific large language model for use as an AI coding assistant by First Light Fusion’s science and engineering teams.

The agreement follows a memorandum of understanding signed by the two UK companies earlier this year.

The model was trained on First Light Fusion’s scientific and engineering workflows using Locai Labs’ proprietary Forget-Me-Not framework. It has been installed on First Light Fusion’s secure computing infrastructure in Oxford, keeping proprietary data and intellectual property within the company’s existing systems.

First Light Fusion works on inertial fusion energy and has also developed machine learning simulation tools for its research. The coding assistant is intended to support work across its science and engineering operations, where specialist software and technical workflows are part of day-to-day development.

The contract also provides for ongoing updates. Under the project plan, the system will undergo two full retraining cycles each year to reflect changes in First Light Fusion’s research activity.

Secure environment

A central feature of the deployment is that the model runs inside First Light Fusion’s own environment rather than through an external service. It has been deployed on secure, air-gapped computing infrastructure designed to isolate sensitive information from outside networks.

That arrangement reflects the sensitivity of data involved in fusion research, where simulation work, experimental results and engineering designs can form part of a company’s intellectual property. By keeping the model within its own systems, First Light Fusion aims to use AI tools without moving research data outside its direct control.

Locai Labs describes itself as a UK artificial intelligence company focused on foundational models. Its Forget-Me-Not method is used to adapt models for specific tasks and organisations, in this case to meet the specialist requirements of fusion energy research rather than general software development.

The contract gives Locai Labs a role in one of the UK’s more technically demanding industrial research fields. Fusion companies increasingly rely on simulation, modelling and data analysis to shorten development timelines and improve experimental design, creating a potential market for tailored AI systems that can operate in secure research settings.

First Light Fusion was founded as a spin-out from the University of Oxford and has focused on technologies linked to inertial fusion energy, defence applications and materials science research. Its infrastructure includes what it describes as Europe’s largest pulsed power facility and the UK’s largest two-stage gas gun.

It has also outlined a strategy called FLARE, which it presents as a route towards commercial inertial fusion energy. The latest agreement suggests the company is also investing in software and internal research tools as it develops that broader technical programme.

Industry context

For Locai Labs, the contract provides a reference point for applying domain-specific language models to industrial and scientific work. Rather than offering a general chatbot, it has supplied a model trained on a single customer’s internal workflows and hosted within that customer’s secure estate.

Mark Thomas, Chief Executive Officer of First Light Fusion, said: “We’ve been impressed by the quality and speed of the work delivered by Locai Labs. From initial planning to the final deployment on our air-gapped system, their team demonstrated a deep understanding of our needs and a commitment to excellence which aligns with our own mission. This AI coding assistant is already proving to be a valuable asset.”

James Drayson, Chief Executive Officer of Locai Labs, said: “This collaboration is a powerful example of how UK-developed foundational AI can be applied to accelerate UK scientific research and industrialisation. It’s a privilege to support such an innovative and mission-driven company.”