Smart Tools for Chemical Research
Iris.ai: Speeding up Research Discovery with Artificial Intelligence
What sounds like having put the cart before the horse turned out to be a brilliant strategy for this company working to have machines make sense of written scientific content for us. According to Fast Company, Iris.ai is the world’s first AI researcher and was named one of the Top 10 Most Innovative Companies in AI in 2017. Michael Reubold met the founders.
CHEManager: Iris.ai was founded 5 years ago. How did it all start and what was there first: the idea or the company?
Anita Schjøll Brede: A balanced and dedicated team of A-players, in our view, is the most crucial and important factor in a start-up, and that was where we started! The fundamental problem we then decided tackle is and was simple enough: we have an unprecedented and exponentially growing body of scientific knowledge, but it is too much for humans to make sense off — so we need smart machine systems to help.
Why did you pick the name of the Greek messenger goddess for your firm?
A. Schjøll Brede: We thought of Iris.ai as a messenger bringing important knowledge from powerful beings — such as researchers — into the right hands, at the right time.
What is the USP or differentiating feature of Iris.ai?
Victor Botev: We allow smart leverage of the knowledge in millions of scientific text documents: without pre-indexing and any other substantial training by humans. This allows our users to do things like: finding new application areas for their existing compounds; extract all key data from a set of documents; build smart knowledge graphs from their own written expertise; and do superior literature searches.
Which obstacles did you have to master so far?
A. Schjøll Brede: Obstacles we have had to master include attracting top-level talent, which we have been able to do both by unifying around a big vision, and by being a remote team where we will hire people wherever they are — especially for Machine Learning engineers, because it allows us to find people who would prefer not to move an AI hub. We’ve also had to work hard to attract VC funding: it becomes extra hard when you receive a lot of attention so everyone wants to talk to you, but you work with complex technology in a complex industry so no one really has you as their core investment thesis. Luckily, we have managed to build an incredible team who stands by us founders in thick and thin, and can pull of absolute miracles when needed!
What have been the most exciting projects so far?
V. Botev: We are right now working with a great client on an extraction tool for key data from patents. It is so satisfying to work in proof of concepts with companies where they know very well what needs they have and what problem they want to solve but recognize that building it themselves is outside the scope of their core expertise. We won the proposal and have been working for 8 months on the project; now we’re getting close to completion and seeing the client happy because they get exactly what they wanted from a very challenging technical task is very exciting.
In a broader perspective, we were recently announced as a Top-10 semifinalist in the “AI for Good” XPRIZE, which we have been competing in for more than four years. It has been amazing to go through the judging stages and see that our research, our technology, its impact, and our team, has been repeatedly vetted and deemed solid, sound and contributing to the field of Artificial Intelligence. It is such an accomplishment by our researchers!
What will be the next steps to develop Iris.ai?
V. Botev: We need to find more of the brilliant and forward leaning chemistry companies to work with and get our smart tools to full scalability! We thoroughly enjoy exploratory discussions with potential new clients.
Anita Schjøll Brede is CEO and Co-Founder at Iris.ai. She is also Faculty in AI at SingularityU Nordic. Forbes claims she’s among the world’s top 50 women in tech. The past decade her career has spanned 10 industries including developing an e-learning tool in Silicon Valley, reducing energy consumption in the process industry, facilitating solar light business in Kenya and trying to disrupt the recruitment industry. She also dropped by six universities on the way. And built a race car.
Victor Botev is the CTO and Co-Founder of Iris.ai, previously a researcher from Chalmers University of Technology. He studied individual master’s degrees in Artificial Intelligence and Computer Systems and Networks at Sofia University St. Kliment Ohridski and Chalmers, respectively. At Chalmers, he conducted research on clustering and predictive neural network models and the usage of signal processing techniques in studying Big Data. He has put his unique combination of AI research, software development lead and ambitious vision to the ultimate test at Iris.ai.
The AI Chemist
Over the past five years, Iris.ai has built an award-winning Artificial Intelligence/Machine Learning engine for scientific text understanding. Now, it has been trained and specialized in chemistry research. That allows the Oslo, Norway-based start-up to build a new suite especially for the chemical industry.
Iris.ai is offering smart tools for Chemical R&D, allowing researchers effortless leveraging of information from millions of documents; research papers, patents, internal documentation etc. The smarter literature search tool ‘Discover’ allows users to visually map out all papers, patents and other documents related to their research question. Teams using this tool are scientifically proven to outperform teams using keyword-based search tools on overview, spot-on papers found, and conclusions drawn.
The ‘Identify’ tool helps users find new applications for their chemicals, materials or compounds, by scouting millions of interdisciplinary, unstructured documents. This systematizing of serendipity is simply not possible with only human brain power and can open up new revenue opportunities for the company.
The ‘Extract’ tool automatically extracts key data from patents, papers and other scientific literature. At 90% accuracy, two full person-months of tedious labor can be automated down to a few hours performed by a machine. The core engine is based on recent years’ breakthroughs in AI, Machine Learning and Natural Language Processing. The algorithms deal with scientific text understanding: Similarity: proprietary WISDM metric measuring how similar two pieces of text are.
- Compositionality: on word level or text piece level, to identify parent/child concepts.
- Causality: advanced knowledge graph connections and basic reasoning to determine causal relationships.
- Corresponding ranking metrics: to give contextually fitted results.
The Iris.ai Chemistry engine can also be used to make custom fitted tools from the modular system.
An Ambitious Team
Iris.ai was founded at Singularity University at NASA Ames Research Park in summer 2015. Challenged to come up with an idea that would positively impact the lives of 1 billion people, the group soon found a common frustration with the access to and passion for the value of scientific knowledge. The founders spent the next four years building a core AI engine and a suite of tools for academic researchers, which has been successfully deployed at University libraries across the Nordics, before taking on the next challenge: specializing the machine on chemistry research and building more pointed tools for industry. All of this leads Iris.ai toward the final goal: «The AI researcher», an AI system that will become an invaluable team member of any human research team in the future.
- Iris.ai was founded at NASA Ames Research Park.
- We were admitted to 500 Start-ups in the Nordics, launched our first simple tools, attracted the first round of angel financing and joined the AI XPRIZE competition.
- Iris.ai was announced by Fast Company to be among the world’s Top-10 most Innovative companies in Artificial Intelligence, the start-up was admitted to Founders Factory in London, and the WISDM paper was published.
- The scalable tool suite for University Libraries was ready, the first full licenses were sold and the Scithon paper was published.
- The decision to focus in on Chemistry was made, the first tools iterated, and the first Proof of Concept clients from the Chemical industry were onboarded.
- Iris.ai was announced as a Top 10 Semifinalist in the AI XPRIZE.
- Sign on an additional limited number of Proof of Concept clients to co-design unique tools.
- Commercial Chemistry tools ready to be scaled with SaaS licences.
- Learnings from Chemistry and other specializations re-generalized, and birth of the first version of the fully fledged AI Researcher.