Intellegens is a spin-out from the world-famous Cavendish Laboratory at the University of Cambridge. The company has developed a unique approach to deep learning from sparse data. Sparse data (as is typical of experimental data) currently poses a big problem for traditional deep learning methods, which are designed to fill in gaps in information. Conventional approaches require reliable, complete data to produce accurate models. However, in the real world, and particularly in R&D, complete data is rarely available.
Intellegens’ flagship product is Alchemite™, an innovative deep learning engine which allows algorithms to be trained on fragmented or incomplete data. The beauty of the engine is that it is generic and can be applied to any numerical domain where sparse data exists. Intellegens is now working with customers in a number of different fields including:
• Drug discovery: looking at protein/molecule interactions in a bid to reduce development costs and the time to market for life-saving drugs and treatments
• Healthcare: looking at patient histories and possible treatment plans to guide physicians; improve the consistency of service and sharing of information; achieve cost savings and deliver better patient outcomes
• Infrastructure: looking at predictive maintenance schedules for the utility industry and how to prioritize work on projects where small improvements to workflow planning could lead to big cost savings, more reliable services, and the more efficient use of resources.
More broadly, Intellegens is also developing a series of software tools, which will allow users to take advantage of its technology in a simple, intuitive way. Working closely with customers to understand more about their needs, Intellegens is developing:
• A simple, web-based tool to build models
• An integrated workflow solution
• A series of pre-generated models to answer specific industry questions; the business is currently working to create an online demonstrator that can accurately predict the physical properties of steel-based alloys based on their composition and response to heat treatment.
Intellegens was co-founded by Dr. Gareth Conduit and Ben Pellegrini. In 2017, Dr. Gareth Conduit joined forces with Ben Pellegrini to help bring his invention to market. Gareth is a Royal Society University Research Fellow in the Theory of Condensed Matter Group, and a Fellow at the Cavendish Laboratory – University of Cambridge. His core research is focused on the design of new materials, and drugs, and quantum phenomena. Ben is an experienced software architect. Previously CTO at a big data startup, Ben is experienced in delivering large data-driven software solutions into corporate, highly regulated business environments.
Dr. Gareth Conduit developed the Alchemite™ engine for deep learning from sparse data while he was designing a new superalloy for use in jet engines – an area of industry where the material design process can run into tens of millions and is incredibly lengthy (between five to ten years). Given the significant amount of historical, materials data available in this sector, and the availability of high-performance computing (HPC), Gareth realised that deep learning methods would be an excellent way of identifying missing correlations and making new material predictions.
The key challenge then was that the historical data available was not uniform or consistent in terms of material property measurement. Traditional predictive methods would need either new models to be generated for each individual material property or, lots of data to be stripped out, which while incomplete could hold additional valuable information. To get around this, Intellegens developed a way of using all the information available from within the datasets – even if it was as little as 0.05% complete. Despite using significantly sparse data, Alchemite™ was able to train models to learn correlations between all available parameters (input or output) and make predictions that were highly accurate. The end result is an intelligent system that enabled the jet engine company in question to identify new materials with properties that could be beneficial when it came to cost, strength, density – without having to invest hundreds of hours and tens of thousands of pounds in expensive R&D and practical experiments.
Intellegens was set up to commercialize the clearly valuable algorithms encapsulated in the Alchemite™ engine developed by Dr. Gareth Conduit. Intellegens was initially focused on applying this technology to improve the advanced materials design process, making it cheaper and quicker for engineering companies to identify and ultimately use more effective materials. Through this process, the Intellegens team has come to appreciate the power of its technology and its potential to be applied across a myriad of industries. The business is now working with key customers across a range of market sectors to understand how its approach to closing the gap on sparse data can be applied to different commercial problems.
Trailblazing Innovation That Empower Industries
Intellegens believes it is the only company in the industry that is able to offer deep learning on such sparse data sets. The company’s technology is still under active development through research at the University of Cambridge, with Intellegens providing a direct route to market and commercial engagement.
Intellegens’ technology is being applied to solve problems in both academic research and multiple commercial projects. Through research channels, academic papers and talks, the business is detailing how its approach to deep learning can solve a number of common data-related problems. Through this work, Intellegens is educating forward-thinking companies, in a range of different sectors, about the opportunities that AI can offer and how this technology can be applied to improve processes, reduce costs and deliver better results.
Big Data and Cloud to Fuel Innovation in AI
“Cheap big data and cloud computing are enabling companies to take advantage of AI and deep learning to identify patterns in data that were previously unattainable. These technologies are supporting all aspects of innovation across industry and society,” said Ben. In the future, the tools and processes used throughout innovation, regardless of application, will become more and more driven by insights from big data, access to cloud computing and the application of algorithms. “In this regard, Intellegens and its deep learning technology will sit at the forefront of developments – helping organizations of all kinds save money by reducing the product innovation/development lifecycle and time to market”.
Awards and Accolades
Intellegens has been honoured with the Innovate UK award in its emerging and enabling technologies category to build a demonstrator for experimental data. This award enabled the first version of Intellegens’ software tool, Alchemite™, to be developed. It is also enabling the creation of a simpler interface that will allow non-expert users to take full advantage of the technology. Cambridge Enterprise also awarded Intellegens a Fast 50 prize to help develop its technology and has been providing the team with hands-on, local support.
Learning Through Challenges
Ben believes the current hype in AI has been a double-edged sword for the team at Intellegens. While the business has a truly unique approach, there is a danger that it will get lost among the numerous companies claiming to use AI, which are actually using off the shelf, inappropriate techniques. The demand for machine learning engineers has also significantly increased, meaning it is much harder to find good candidates. Access to data is another challenge that Intellegens faces. Data is increasingly seen as an organization’s intellectual property (IP) and therefore getting access to it can prove difficult. Intellegens has overcome this by providing tools that can be deployed in enterprise environments, on individual computers, or through the cloud, giving customers peace of mind that their information is secure.
Building Solutions for Tomorrow
Intellegens is at an early stage of its development but has an exceptional team alongside a unique set of tools for understanding and solving the many information-based problems that businesses worldwide run into every day. Working closely with customers, Intellegens is still learning about how its technology can be tailored and applied to work across multiple sectors. The possibilities are potentially endless. With a number of high profile projects nearing completion, the business is looking forward to the next stage of growth, building on existing relationships and projects, and identifying new opportunities.