Canada-based startup Infera AI seeks to bridge the gap between hardware manufacturing and cutting-edge artificial intelligence (AI) technologies. With their bespoke solutions and subject-matter expertise, Infera AI offers data-efficient AI platforms for various industrial applications. In an interview with DIGITIMES Asia, Infera AI co-founder and CEO Trefor Evans discussed the company's background, vision and where it stands in the future.
Evans brings a wealth of experience to Infera AI, with a background that combines engineering and machine learning. As a founding member of AeroVelo Inc., he led the structural design and optimization of a human-powered helicopter that won the prestigious Igor I. Sikorsky prize in 2013.
Evans further honed his expertise in machine learning by completing his PhD at the University of Toronto. During this time, he served as the lead developer of a machine learning library used in jet-engine development programs at Pratt & Whitney Canada, a global leader in gas turbine manufacturing. These experiences solidified his belief in the immense advantages AI can bring to the engineering and manufacturing sectors.
"Infera AI was founded as I completed my PhD to build exceptionally reliable and data-efficient digital twins – a digital representation of a physical asset or process," said the CEO. "Digital twins can be applied in so many different sectors, and we've pivoted a couple of times throughout the life of the company." According to Evans, Infera AI was originally built to focus on the engineering sector – particularly engineering design process - but experienced some challenges. "Our technology for assisting engineering design required a software infrastructure that was not mature enough in most industries," noted Evans.
Since then, the company recognized the opportunities in the manufacturing sector, including quality assurance monitoring of manufactured parts using virtual metrology, predictive maintenance of equipment & assets, process optimization & control for manufacturing.
Infera AI distinguishes itself by adopting a consultative and collaborative approach to streamline the process of transforming an industry problem into a machine learning problem, solving it with a statistically rigorous machine learning approach, and delivering tailor-made AI platforms for deployment. Years of leading AI research have culminated in Infera's proprietary AI platform which performs fully probabilistic inference to capture uncertainty and allows easily incorporating subject-matter expertise, constraints, and physics. This approach allows the company to operate with significantly less data than its competitors, which is particularly crucial in cold-start scenarios when setting up new processes. As a result, their predictions are not only sensible but also more confident and accurate.
Evans notes that though Infera AI is a software company, its expertise ultimately lies in bridging the gap between hardware manufacturing and the state of the art in AI. "The greatest impacts of industry 4.0 will be seen when there are manufacturing subject-matter experts working closely with AI-experts. And this is precisely how we have structured our company," noted Evans.
As part of its expansion plan, Infera AI is currently identifying a target vertical to expand within. In the long run, the company aims to localize their technologies for various verticals and customize their AI platform accordingly.
Meanwhile, the company is actively seeking partnerships with Taiwanese manufacturers, especially those in the semiconductor sector, interested in piloting their AI platform for quality assurance, process optimization and control, predictive maintenance and monitoring, or data analytics for business performance.
"To give an example where our AI platform can be used, we can use it alongside an ion beam implantation process which is a fundamental process in semiconductor fabrication," noted Evans. The CEO indicated that Infera's platform has been demonstrated in the quality assessment of dosed 300mm wafers subject to several important constraints, such as dosing uniformity, using the company's virtual metrology capabilities.
In addition, Infera has also applied the platform's process optimization and control capabilities to calibrate and optimize the dose profile provided by its AI platform in order to reduce tool downtime and wasted materials. "Using a simulation based on several real implantation tools, we demonstrated the reduction of material waste by 48% and the delivery of a greater total implantation dose using the same total implantation time and uniformity requirements compared to the existing industry methods."
Jialian Wu (co-founder and chief product officer) and Dr. Trefor Evans (co-founder and CEO)
