Orion

Orion, a globally operating pharmaceutical company from Finland, wants to supercharge research and development of innovative medicines. The onset of large language models (LLMs) such as GPT-4 creates new opportunities to summarize and understand research information and insights at record speeds.

Partnering with Reaktor, a long-standing collaborator for Orion, they set out to do just that. In just four months, we built a standalone tool that could summarize extensive research documents, presentations, and other material, respond to specific and generic inquiries, and significantly reduce the time spent on document and insight review.

Pharmaceutical research is notoriously meticulous and time-consuming. To gather insight, researchers write and review many lengthy and complex documents and presentations, often as long as 300 pages or more. Additionally, a certain topic or notion of interest can span dozens or even hundreds of such documents and papers. Sifting through that amount of data to extract relevant information or answer specific questions is a laborious and time-intensive task. 

And it doesn’t stop at the drug research phase. The medical affairs department, responsible for creating a thesis landscape prior to a commercial launch, goes through vast amounts of material in order to get a comprehensive understanding of patient journeys, patient needs, the competitive landscape, available treatments, and more.

This work often begins many years before the commercial launch of a drug. Unlike sales teams, medical affairs can engage in understanding the competitive and commercial landscapes, guidelines, and treatment protocols across multiple regions and regulatory systems before the drug is even approved to enter the market. By analyzing current and emerging treatments, understanding market predictions, and identifying gaps in different geographical areas, medical affairs teams help tailor strategies for the drug’s launch and its subsequent commercialization.

Orion was looking to develop a ChatGPT-like tool, but tailor it specifically to the challenges a pharmaceutical company would naturally face, particularly when it came to the work of their drug development researchers and the medical affairs team. Most importantly, Orion needed to make sure the tool could provide the ability to process vast amounts of documents while ensuring the answers are factual by providing references to the original text via citations.

We kicked off the project with a light discovery phase, speaking to key stakeholders to align on needs and requirements. This phase made it clear that there was a need for an augmented search capability that could understand and contextualize nuanced questions while providing a more accurate and detailed response than a traditional search engine would.

As is standard practice for us, we broke the project down into pieces and took an iterative approach to development. In six weeks, we had a functioning PoC that gave us the green light and confidence to proceed towards an MVP.

We built a standalone tool that operates similarly to ChatGPT but is tailored to the specific needs of the pharmaceutical industry. The tool allows researchers to create projects, upload multiple documents, and interact with the content through a chat interface. The key features included:

• Project-based Interaction: Users can upload numerous documents and files within a project, helping with organized research and easy retrieval.
• Advanced Text Recognition: Recognizing and extracting text from complex layouts and files.
• Patent Analysis: The tool extends to patent documents, which are crucial for evaluating competitive or collaborative opportunities.
• Ease-of-use: An important factor in driving wider organizational adoption of the tool was to pay special attention to design with a simple, modern, and user-friendly interface.

The most important feature of this tool was the work we did around factuality and source verification. To combat misinformation and AI hallucinations – a common reliability concern with AI-generated content – the tool incorporates mechanisms to verify facts and cite original sources. This provides a level of factuality assurance and easy access to double-check the content’s accuracy. 

Adopting this AI-powered tool has significantly accelerated both the research phase in the drug development process as well as helping the medical affairs team to build a thesis landscape more efficiently. Increasing efficiency in these areas enables Orion to move faster in its innovation efforts. The initial success has sparked interest across the organization in further exploring and expanding the use of Gen AI technologies.

• UI & UX design
• Web development
• Data Science

Orion’s research ecosystem leverages data and AI to accelerate pharmaceutical research and development, now and in the future. This collaborative network is made up of a variety of Finnish, European and global companies, hospitals, universities, and research institutes, and focuses on exploiting data and novel in silico models to reduce drug development times. This project was done under the realm of Orion’s research ecosystem in partnership with Reaktor.

As a leading pharmaceutical company in the Nordics, Orion is always on the lookout for top research collaborators to develop new products, services, and solutions for the healthcare market. 

If that is you, please visit https://www.orion.fi/en/ecosystem and contact Orion directly.