Periodic Reporting for period 1 - The Digital R&D team member (A Digital R&D team member, automating scientific knowledge handling, allowing European R&D to increase their speed of innovation)
Okres sprawozdawczy: 2023-05-01 do 2024-04-30
At Iris.ai we have a unique perspective. We believe that if one person could absorb all of the world’s scientific knowledge, they could solve many of humanity’s challenges. But our human capacity is limited. That's why we founded Iris.ai to build an AI that can make sense of and connect the dots in this vast knowledge. We are not just another AI company, we are an impact-driven company with a unique mission.
From a more commercial perspective, our clients and users' pain points are the intersection between an exponentially growing body of human scientific knowledge, increasing pressure to stay competitive through R&D and innovation efforts, and a realisation that while most other industries and departments are receiving digital makeovers, R&D is unusually stagnant in its ways. Our clients are R&D-heavy organisations in the chemical, pharmaceutical, MedTech, biotech, and material science industries. Their core operations depend highly on scientific content, from papers and patents to internally produced proprietary research. They have heavily staffed teams where most contributors will have education at least at the PhD level and can spend up to 40% of their time searching for aggregating and cleansing data. With a rapidly increasing body of content (6,000 papers and 8,000 patents a day, growing at 5% p.a.) their problem of accessing the proper knowledge at the right time to push innovation forward is becoming insurmountable. These companies operate in industries with heavy competition and depend on their scientific breakthroughs to stay ahead of their competitors. This hard-to-solve conundrum is causing a significant problem for these players. Navigating all these knowledge issues falls into two categories: One, they can or are required to do incredibly time-consuming tasks. For example, systematically reviewing new literature and extracting key findings. Two, tasks that are so time-consuming they cannot do them at scale even when they know it would yield significant value. Tasks include searching for and extracting detailed scientific information from competitor patents to gain a competitive advantage or covering all ground needed when developing new products. No matter the fundamental premise of their problem, the activities they want to, need to, or wish they could undertake falls into a basic category: Unlock insights from documented research.
Our current tools are just the beginning. We have a grand vision for the future of Iris.ai. We are working towards creating a fully-fledged scientific assistant, a tool that can answer questions about data sets, guide users through their processes, ask the right questions, bring novel insights or interdisciplinary knowledge, and unlock discoveries. This is not just a dream. It's our roadmap for the future.
From a more commercial perspective, our clients and users' pain points are the intersection between an exponentially growing body of human scientific knowledge, increasing pressure to stay competitive through R&D and innovation efforts, and a realisation that while most other industries and departments are receiving digital makeovers, R&D is unusually stagnant in its ways. Our clients are R&D-heavy organisations in the chemical, pharmaceutical, MedTech, biotech, and material science industries. Their core operations depend highly on scientific content, from papers and patents to internally produced proprietary research. They have heavily staffed teams where most contributors will have education at least at the PhD level and can spend up to 40% of their time searching for aggregating and cleansing data. With a rapidly increasing body of content (6,000 papers and 8,000 patents a day, growing at 5% p.a.) their problem of accessing the proper knowledge at the right time to push innovation forward is becoming insurmountable. These companies operate in industries with heavy competition and depend on their scientific breakthroughs to stay ahead of their competitors. This hard-to-solve conundrum is causing a significant problem for these players. Navigating all these knowledge issues falls into two categories: One, they can or are required to do incredibly time-consuming tasks. For example, systematically reviewing new literature and extracting key findings. Two, tasks that are so time-consuming they cannot do them at scale even when they know it would yield significant value. Tasks include searching for and extracting detailed scientific information from competitor patents to gain a competitive advantage or covering all ground needed when developing new products. No matter the fundamental premise of their problem, the activities they want to, need to, or wish they could undertake falls into a basic category: Unlock insights from documented research.
Our current tools are just the beginning. We have a grand vision for the future of Iris.ai. We are working towards creating a fully-fledged scientific assistant, a tool that can answer questions about data sets, guide users through their processes, ask the right questions, bring novel insights or interdisciplinary knowledge, and unlock discoveries. This is not just a dream. It's our roadmap for the future.
Our tools today, collected in the Researcher Workspace, enable the extraction task outlined above and tasks like 1) getting an overview of your unique research documents. The users are fully in control of what content they use, and on that content, they can rapidly apply tools to understand what they have in front of them. 2) Focus on the right content with machine aid. Users can narrow their document sets from millions to precisely what they need with smart tools going above and beyond keywords. 3) Draw out insights through interrogation. The days of skimming page after page for the right information are gone. Users can interrogate their data set to unlock the insights they need. 4) Lean back to wait for updates. With our tools, vocabulary changes over time is no challenge. Automatic monitoring for new publications has never been easier.
Over the past year, we have successfully built this from scattered tools and a range of TRL6 technologies into a fully-fledged market-leading research platform. The Workspace is content agnostic, and clients can load, add, or connect various knowledge sources - external and internal. We provide full privacy and security and do not use our clients’ content, data, or inputs for anything but their process, which is a must-have that few competitors have. Research demands rigorous attention to facts and no room for hallucinations, and we go beyond LLMs to enforce the ground truth. We focus on optimal human-machine collaboration, but when something is repeatable enough to be automated, it can be automated. We use smart, not large, language models where they suffice. Better for pricing and the environment. Finally, the ML models are rapidly reinforced in the client’s research field to handle their unique vocabulary context. This means the solution is rapidly scalable to various challenges, from core R&D via medical regulatory affairs to IP analysis.
Over the past year, we have successfully built this from scattered tools and a range of TRL6 technologies into a fully-fledged market-leading research platform. The Workspace is content agnostic, and clients can load, add, or connect various knowledge sources - external and internal. We provide full privacy and security and do not use our clients’ content, data, or inputs for anything but their process, which is a must-have that few competitors have. Research demands rigorous attention to facts and no room for hallucinations, and we go beyond LLMs to enforce the ground truth. We focus on optimal human-machine collaboration, but when something is repeatable enough to be automated, it can be automated. We use smart, not large, language models where they suffice. Better for pricing and the environment. Finally, the ML models are rapidly reinforced in the client’s research field to handle their unique vocabulary context. This means the solution is rapidly scalable to various challenges, from core R&D via medical regulatory affairs to IP analysis.
A component of our project has been the introduction of the Researcher Workspace to four pilot clients. These pilot deployments have been initiated and, while not completed, have already demonstrated the platform’s flexibility and effectiveness in various research settings. The pilots aim to highlight several vital outcomes: successful implementation in diverse environments, identification of new use cases within each organisation, and effective training and feedback mechanisms are off to a good start. Additionally, the already consistent request for new user features indicates high engagement and has informed Iris.ai’s product roadmap both for the duration of the pilots and beyond. The deployments also revealed the potential for expansion and scaling, suggesting that Iris.ai could be applied across broader segments within these organisations. Beyond meeting immediate research needs, these pilots serve as strategic tools for deepening client relationships and fostering longer-term partnerships. The findings indicate that the Iris.ai Researcher Workspace can enhance research processes, adapt to various demands, and promote user-driven innovation while aligning with each organisation’s strategic goals.
In addition to bringing these tools to enterprise readiness and executing the pilots, both are works in progress. We are working actively to create our pipeline for commercial development, a key project component, and secure longer-term funding for our growth phase. Both processes are well underway.
In addition to bringing these tools to enterprise readiness and executing the pilots, both are works in progress. We are working actively to create our pipeline for commercial development, a key project component, and secure longer-term funding for our growth phase. Both processes are well underway.