Building for Healthcare: Our First AI Hackathon

Hackathons are a brilliant way to demonstrate the power of multidisciplinary teams combined with a collective goal. At Doctolib, we strive to build the healthcare we all dream of, and we see AI as a key asset to leverage. Naturally, our first hackathon was centered around finding innovative solutions for our industry — and it was a lot of fun!

Here’s what we learned organizing it, how it positively impacted our teams, and which solutions stood out.

A Unique Event to Deliver Impactful Solutions

In the last few months, we organized many training sessions for our teams to get them comfortable with new AI technologies. These sessions were led both by internal and external experts. We imagined this event as a way for our teams to try AI tools, leveraging their transformative power for healthcare. It was the culmination point of an ambitious AI Bootcamp for all Tech teams to put in practice what teams have learned over the past months.

We held the first Doctolib AI Hackathon on July 10–11, and it was a moment of innovation, excitement, creativity and collaboration. Picture this: 137 participants working together in 17 teams across France, Italy, Germany and the Netherlands worked hard to prototype future transformative solutions for healthcare professionals and patients. This first hybrid edition, taking place in our offices with some people working remotely, combined several teams and roles from different countries — software engineers, data scientists, engineering and product managers, designers, and more.

What made this event unique was integrating healthcare professionals in it. Usually, about ten healthcare professionals are present weekly in our offices to answer questions, support teams in creating useful solutions and help us deliver more impact. For this hackathon, we wanted to build solutions to solve real healthcare problems, so five of them were supporting teams, bringing their expertise and unique point of view.

“I really like the idea and all the support we had. Having available healthcare professionals was very helpful and motivating.”

Winning Teams

We had 3 prizes in our hackathon. Here is a summary of our winning teams, their respective concepts, and how we evaluated their impact.

Our jury of 6 people, including the CEO, CTO, a healthcare professional, and members of higher management, had to evaluate (give points from 1–20) each team considering the following criteria:

1. Impact and potential for business application
2. User experience and design quality
3. Innovation and creativity in applying AI technologies
4. Technical implementation and use of data
5. Presentation and communication of the project

The AI Trailblazer, rewarding the most impactful & innovative solution

For this category we had 3 winners, just like in the Olympics.

In 1st place: Team “Sophia” — The Future of Clinical Decision Support
This project aims to provide healthcare professionals with full clinical support during consultations, suggesting real time follow-up questions based on what is being said, offering recommendations, and preparing prescriptions to send to the patient. The model is continuously updated with the latest medical knowledge to ensure reliability. A true game changer!

In 2nd place: Team “Park side” — Health task manager
In a world where healthcare costs and chronic diseases are on the rise, prevention becomes key. But implementing it successfully and durably can be complex. To make it work, this team imagined an ultra-personalized and very engaging platform making prevention a healthy daily fun habit.

In 3rd place: Team “Life in a Doc” — Scan Document and extract key information
For a healthcare professional or a medical secretary, typing information in reports and documents takes away time that could be used for more impactful tasks. This team introduced the “Docto Scan”, to extract meaningful information from paper documents, to store and structure digitally in a matter of seconds.

The Data Wizard, rewarding the best use of data

The winner of this award was the team who got the most points in the two following criteria:

1. Innovation and creativity in applying AI technologies
2. Technical implementation and use of data

Team DoctoCompanion — An agent that automates actions for the practitioner
What would the AI assistant for healthcare professionals look like in the future? This team imagined a powerful voice assistant to support before, during and after consultations.

The Showstopper, rewarding the most entertaining presentation

The winner of this award has been the team who get the most points in the “Presentation and communication of the project” criteria.

Team “DoctoSpotlight” — Reduce access time to features
Imagine: you’re a healthcare professional and you suddenly have a question about the Doctolib solution you are using. This team imagined a tool to answer any question — providing the right information and suggesting the next best action to take.

Kudos to each team, who brought so many ideas to the table!

Participant’s feedback: a great trial & error process

What participants preferred in the event was the spirit of curiosity of teams and the openness to explore AI technologies. Many teams highlighted in their live demos the challenges and the process of trial and error they met while conceiving their solutions. This transparency was well received, as it provided great insight into the process which was as interesting as the prototype presented.

Another aspect of the hackathon that really resonated with participants was working with very diverse teams and experts up close, in designing a practical solution, while having access to available healthcare professionals to answer any questions they might have.

“I really liked working directly with devs as a data scientist. And just in general, super cool to work with new people in a really intense side by side manner.”

Behind the Scenes: Organizing the Hackathon

What Made This Event Successful

Managing this event was full of challenges: we had to prepare a hackathon set in a hybrid mode, with 4 different locations, adding complexity to its coordination and organization. Maintaining a high energy in all locations proved to be another challenge.

Here are our key takeaways on what made this first edition a success:

• We had a strong sponsorship from senior leadership and the Data Teams to set the frame/vision, constitute the jury and support on the tooling (datasets, LLMs, etc.),
• The top management decided to stop the train for a couple of days in order for the participants to truly enjoy and focus on their projects (no slacks/email/meetings),
• To ensure clarity for all, we built a strong and dense communication plan to keep the momentum going as it was very challenging to engage teams that are managing multiple daily priorities in addition with the summer break around the corner,
• We developed a strong learning culture within teams that are willing to get out of their comfort zone and put their knowledge into practice.

Tools used: GPT-4o led the way

For the AI Hackathon teams teams had the privilege of utilizing cutting-edge technologies from Azure OpenAI, including GPT-3.5, GPT-4, GPT-4o, Ada-002, and DALL-E-3. This event provided a unique platform for participants to harness these advanced AI models, driving innovation and pushing the boundaries of natural language processing and image generation.

The hackathon saw an astounding total of 107.29 million prompt tokens processed, a total of 89,490 calls were made to the secure LLMs endpoints developed internally, reflecting the dynamic and productive environment of the hackathon. Notably, 57% of these calls were directed to Ada-002, highlighting its efficiency and popularity for specialized tasks. Meanwhile, 40% of the calls were made to GPT-4o, showcasing its optimized performance and versatility in handling complex queries.

Writing the future of healthtech

Overall, we were all impressed by the way the projects were structured and focused on impact. We have a strong culture of building useful solutions, pushing the envelope for our users. We’re strengthening this aspect of our culture even more now, making sure we form one team with our users, are bold in our innovations, and deliver solutions with a scientific approach.

These innovative solutions tackle some of the most pressing challenges in healthcare, including time constraints, and limited resources for practitioners. By aligning with our OKRs, we are thrilled to integrate the top ideas — the winners — into our roadmaps. With their inclusion, we are excited and optimistic about a brighter future for healthcare!

Key takeaways from the event

• Learning: take the time to estimate the learning curve when it comes to AI tools and implement your training strategy accordingly. We made sure to build a step-by-step learning journey for our teams, with the hackathon as its culminating practical challenge. This helped people feel comfortable in using AI tools, raised their confidence level, and motivated them to join in,
• Culture: make sure to foster a culture of openness and learning, to put participants in the best possible set up to invent and create things together. We introduced the hackathon as a practical part of an extensive training program on AI, where trial and error was accepted,
• Logistics: don’t overlook them, especially if like us, you’re dealing with a hybrid setup and multiple locations! Make sure you’ve got an extensive communication plan to inform each team at the same level,
• Impact: integrate your key users to the organization of the hackathon. For us, it was health professionals, as we wanted to make sure our solutions were truly helpful,
• Sponsorship: One clear way to showcase the importance of a topic is to have your top management involved, sending a clear message to teams.

If you’re passionate about building the healthcare we all dream of, check our available roles.