Newly funded AI startup joins 35-year-old research centre to test efficient models in real industrial environments
Refiant AI, an AI/ML company that compresses frontier models to run at up to 100x lower compute and energy cost, has partnered the Sargent Centre for Process Systems Engineering, a joint research centre between Imperial College London and University College London.
The move represents a major step for Refiant AI, which announced its first funding round earlier this month. Refiant AI joins a consortium of leading industrial players, placing it amongst some of the most reputable companies in the world.
Ultra-efficient AI, at the edge
Today’s most powerful AI models require enormous amounts of energy and computing power to run – typically through cloud data centres controlled by a handful of technology giants.
For industries that handle sensitive operational data or need AI to work on-site in real time, that dependency creates problems of cost, privacy and reliability. These sectors need models that can run locally or in the cloud, handle large document workflows, and deliver results without compromising on accuracy.
The collaboration will focus on investigating whether compressed, energy-efficient AI can meet the demanding requirements of the process industries. Refiant AI has previously reported the ability to compress a 120-billion parameter model to run on a standard laptop with 12GB of RAM, retaining 95-99% of the original model’s accuracy and reasoning capability. At a time when the largest AI companies are spending tens of billions on GPU clusters and data centre buildouts, Refiant takes a different approach – delivering the same performance with a fraction of the compute.
35 years of engineering firsts
Refiant AI will work closely with the Sargent Centre’s Autonomous Industrial Systems Lab (AISL), which focuses on building intelligent systems that can operate with minimal human intervention. Together, they will develop a standardised framework for testing whether AI systems can understand cause and effect in real industrial settings – and whether they can support operators making high-stakes decisions safely, quickly and locally. The research will be peer-reviewed and published.
The Sargent Centre has operated for over 35 years and is a highly successful collaboration between Imperial and UCL, leading London universities, and has produced several spin-out companies, one of which received Royal Academy of Engineering MacRobert Award – the UK’s highest honour for engineering innovation.
“The AI industry has a scaling problem that it keeps trying to solve with more money and more hardware,” said Dr. Viroshan Naicker, Co-Founder of Refiant AI, and PhD Mathematician. “We went the other direction. We compress models so they run wherever they are needed – on local machines, at the edge, without sending data to the cloud. Working with Imperial College London, University College London and the Sargent Centre gives us a chance to validate our approach in some of the most demanding operational environments on earth.”
“Running powerful AI models reliably and with speed at the edge is a prerequisite for autonomous industrial systems” said Dr Mehmet Mercangoz, ABB Associate Professor in Autonomous Industrial Systems, Sargent Centre.
Looking forward
The partnership marks a period of rapid momentum for Refiant AI, following its $5 million seed round led by VoLo Earth Ventures. The company’s platform uses nature-inspired algorithms to compress AI models while preserving their accuracy. Beyond compression, the company is developing advanced capabilities in ultra-long context management. Early results from both are expected to be shared soon.
