Doctolib : Déployer une Stratégie IA Générative 🚀🤖 (#127)

Summary

In this engaging podcast episode, Nassim, Senior Director of Data and AI at Doctolib, discusses the challenges and strategies involved in implementing generative AI within the healthcare industry. The focus is on the transformative potential of AI tools like ChatGPT in medical settings, from easing administrative burdens for practitioners to enhancing patient care through advanced documentation and medical record handling. The episode also covers the technical aspects of integrating AI technologies into existing systems, the rapid evolution of AI capabilities, and Doctolib’s proactive approach in leveraging AI to revolutionize healthcare services.

Takeaways

• 🚀 Nassim, Senior Director of Data and AI at Doctolib, discusses deploying generative AI strategies, highlighting recent developments due to the influence of technologies like ChatGPT.
• 👥 Doctolib's medical assistant tool is a collection of use cases designed to streamline operations for medical practitioners by leveraging generative AI for tasks like note-taking during consultations.
• 💻 The team includes founding members and the CEO, demonstrating significant leadership involvement and a direct impact on the project's advancement.
• 📈 Rapid advancements in AI technology pose challenges as the 'truth of today' may quickly become outdated, necessitating a highly adaptable and flexible approach to project management and implementation.
• 🔧 The introduction of an internal tool named 'Docto GPT' allows Doctolib staff to access LLM securely, fostering familiarity with the tech and encouraging innovative uses within the company.
• 🌐 The Working Group formed to explore AI applications at Doctolib includes a diverse mix of roles—from engineering to legal—ensuring comprehensive coverage of ethical, technical, and practical aspects.
• 📊 Real-world applications of the AI tool at Doctolib include creating more efficient documentation workflows and potentially enhancing diagnostic processes, illustrating the practical benefits of AI in healthcare.
• 🔍 Challenges in implementing AI involve managing costs, optimizing GPU usage for model training and inference, and integrating new AI technologies with existing data infrastructure.
• 👨‍⚕️ The medical assistant AI focuses on reducing administrative burdens, allowing practitioners to spend more time on patient care and improving the overall quality of medical documentation.
• 🔄 Nassim emphasizes the importance of continuous learning and adapting within AI projects to keep pace with rapidly evolving technologies and to maximize their potential impact.

Q & A

What is the main focus of the medical assistant project at Doctolib?

The medical assistant project at Doctolib is focused on creating a collection of use cases that facilitate the work of medical practitioners. It aims to generate consultation notes from the interaction between the doctor and the patient, thereby reducing administrative tasks and allowing doctors to focus more on patient care.

How does Doctolib's CEO contribute to the AI strategy?

The CEO of Doctolib is very directly involved in the AI strategy, showing a high level of commitment and regular engagement with the project. He has spent time understanding the intricacies of neural networks and deep learning models, demonstrating a personal enthusiasm that aligns with the company's strategic direction.

What is the role of the Working Group in the development of AI at Doctolib?

The Working Group at Doctolib was created to understand the use of Large Language Models (LLMs) within their tech stack. It aimed to identify potential use cases for LLMs within the company and has evolved to prioritize use cases that have a significant impact on users while considering the associated risks, especially in the medical field.

How does Doctolib ensure the secure and effective use of LLMs by its employees?

Doctolib developed an internal tool called Doctolib GPT, which is based on foundational models and allows any Doctolib employee to securely access an LLM. It enables users to choose different types of LLMs for various use cases and to perform prompt engineering in a user-friendly manner.

What are some of the use cases that have emerged from the AI strategy at Doctolib?

Some of the use cases include generating consultation notes from medical appointments, creating referral letters to specialists, and leveraging LLMs to improve the documentation process for doctors, which can enhance the quality of patient care and free up time for doctors.

How does Nassim, the Senior Director of Data and AI at Doctolib, perceive the challenges in deploying a generative AI strategy?

Nassim sees the rapid pace of change as a significant challenge, noting that the 'truth' of today can be invalidated weeks later. This requires a highly flexible and agile approach to stay adaptable and iterate quickly based on new insights and developments in the field of AI.

What is Nassim's background prior to joining Doctolib?

Nassim worked for 15 years in the travel industry, with about half of his career in software engineering and the other half in data-related fields, including data science, analytics, data engineering, and traditional BI. He managed the reservation engine for a part of Expedia and later transitioned into the world of data.

How has the COVID-19 pandemic impacted Nassim's previous work in the travel industry?

The COVID-19 pandemic had a significant impact on the travel industry, leading to less favorable periods with fewer opportunities for innovation. The part of Expedia where Nassim worked was acquired, and he foresaw several years of system integration work, which he had already experienced in his career.

What motivated Nassim to transition from the travel industry to Doctolib?

Nassim was motivated by the desire to have a direct impact on users and patients and to contribute to the healthcare system. He was attracted by the opportunities at Doctolib and the chance to work on projects that would directly benefit users, rather than focusing on internal constraints or integration projects.

What is Doctolib's approach to integrating AI into its products and services?

Doctolib has taken a holistic approach, considering the entire lifecycle of a doctor's day and identifying areas where AI can add value. They have focused on use cases that solve real business problems and have a significant impact on users, while also considering potential risks, especially in the medical context.

How does Doctolib ensure that its use of AI aligns with its broader company strategy?

Doctolib ensures alignment by involving representatives from various departments, including engineering, product, data science, legal, and ethics, in the development of its AI strategy. This cross-functional approach helps to consider the implications of AI from multiple perspectives and ensure that the technology is used responsibly and effectively.