Yuliia Hladka


How Life Sciences companies can start with AI today

If you’d like to learn:

  1. How companies in Life Sciences can create value with AI
  2. How AI assistants can reach superhuman accuracy for certain tasks in drug and vaccine development
  3. How AI can go beyond drug discovery to solve many low-effort but high-impact problems in the drug development process

… then read on!

Our CTO Laurent Sorber participated in a digital session hosted by EyeForPharma, where he presented how Radix built an AI solution for GSK, one of the world’s largest pharmaceutical companies. (P.S. If you missed it live, you can catch his presentation here, and then come back to this article: https://youtu.be/eCo0m3IwdR0.)

Here are the 6 questions Laurent got from the audience, a mix of scientific and commercial profiles:

1: What would be your top 3 tips for a Life Science company to get started with AI?


  1. Identify the opportunities where AI can contribute most. The best places to apply AI are those where you feel you’re not very efficient, or where the quality of the current output isn’t high enough, or where the cost is too high.
  2. Make sure to define how to measure the impact of the solution. By doing that, you will be able to identify both AI and non-AI opportunities to maximize the impact throughout the project.
  3. Make sure you try to get a working version to your end-users as soon as possible. Once you show it to your end-users (e.g., lab technicians, statisticians etc.) they can provide valuable feedback that you can iterate on until you reach an optimal solution that maximises the value for both the business and end-users.

2: How long does it normally take from the ideation phase to a finished solution like the one for GSK?

Laurent: We need about 2–4 weeks (1–2 sprints) to go from ideation to a first working solution. From there, end-users can provide feedback on what the next most valuable areas are for us to work on. Each sprint thereafter we focus on those tasks that maximise the value for the business and end-users. In the case of the AI-assistant for CFU counting we developed for GSK, we needed about 6 sprints in total to arrive at the final product.

3: Do you need to have a use case in mind to get started with AI?

Laurent: You may already have a project in mind or an idea of the impact you want to achieve. If that’s not the case, we have created a Fast Discovery workshop with which we help companies identify the most valuable opportunities for AI. In this workshop, we use an approach based on design thinking to connect those areas in your business and processes where AI can contribute the most. At the end of this workshop, the output is a list of candidate projects ranked by least effort and most impact that are ready for approval by a project sponsor. 

4: How do you deploy and scale your solutions?

Laurent: The project was intended as a Proof-of-Concept (PoC). That being said, we developed the application from the ground up so that it could be easily deployed into a production environment. It’s fairly easy to achieve something that is close to production-ready in the PoC stage already by dockerizing the solution — packaging it up as a Docker image. This is what we did for this application: the docker images that we developed run on GSK’s Azure cloud platform. As we speak, their system administrators are looking into scaling this out for use in production for a first pilot lab. From there, they will be scaling it out to other labs across GSK as well.

5: Since you achieved superhuman accuracy, won’t lab technicians be substituted by an AI over time and lose their jobs?

Laurent: Not at all! Everywhere where we’ve applied AI so far we see that AI is actually positioned as an assistant for humans to enable them to be more effective at their jobs. Because lab technicians will have more time on their hands now, they may focus on more challenging and rewarding tasks that require human creativity and problem-solving skills. Many processes in Life Sciences could benefit from AI’s ability to make these more efficient, with better outcomes and at lower costs. For example, automating the reporting on (pre-)clinical studies and the assembly of initial applications have the potential to be partially or completely automated with the help of AI.

6: Are you applying AI in clinical research that isn’t related to image processing?

Laurent: Yes, we also work on other projects in Life Sciences, and one of them involves Natural Language Processing. There are enormous amounts of documents written, processed, reported, and collected in Life Sciences — that’s one large area where AI can help.

If you’ve missed Laurent’s presentation on how Radix helped lab technicians save up to 6x in one of the pharmaceutical production processes, watch a recording here: https://youtu.be/eCo0m3IwdR0

For more on AI in Life Sciences and the way we create impact with our solutions, please visit our Cases section.