Uploaded June 2025 | Updated September 2026, 2 weeks ago
#datascience #genAI #clinicaltrials #clinicalreporting #AIinpharma #GenerativeAI #Roche #AIautomation #pharmaceuticalindustry #AIchallenges #AIbot #chatbots #copilot #codingassistant
In this episode, we speak with Vincent Shen, Senior Principal Data Scientist at Roche, about the actual impact of generative AI in clinical data analysis. Vincent shares insights from his work implementing AI-powered tools, such as knowledge chatbots and coding assistants, that support clinical reporting at Roche. We discuss the evolving role of AI, the challenges of integrating it into clinical trial workflows, and the importance of rethinking the processes to ensure effective AI use.
In this episode::
- AI will augment rather than replace data scientists in clinical trial workflows.
- Roche has successfully deployed a GenAI-powered knowledge bot, saving an estimated 25 hours per 100 questions.
- Coding assistants help data science teams accelerate time-to-insight.
- Providing detailed specifications is essential for AI to deliver accurate and meaningful results.
- Training AI models requires extra effort to ensure they correctly interpret context and domain-specific requirements.
- The next big step is to rethink how processes are designed, not just automate existing ones.
Vincent Shen
linkedin.com/in/vincent-shen-vs
Nat Chrzanowska
linkedin.com/in/nat-chrzanowska
__________________________________________
More about Appsilon: ► appsilon.com
Appsilon empowers pharmaceutical and life sciences companies to leverage open-source technology for faster, data-driven decision-making in regulated environments. Schedule a free consultation with our expert
► appsilon.com/contact-us
We design scalable and user-friendly Shiny dashboards to help you make data-driven decisions.
► appsilon.com/services/data-dashboards
We design and implement Statistical Computing Environment for R and Python for efficient data analysis:
► appsilon.com/services/sce
We design, implement, and optimise data analysis environments so you can focus focus on insights and innovation.
► appsilon.com/services/platform
Where Technology Meets Science podcast is available on all podcasting platforms:
► Spotify: https://hubs.li/Q037KddD0
► Apple Podcasts: https://hubs.li/Q037KlLl0
► YouTube: https://hubs.li/Q037Kpth0
__________________________________________
For more insights about how technology helps scientists push the boundaries of data analysis and reporting check out our blog:
► appsilon.com/blog
LinkedIn: linkedin.com/company/appsilon
__________________________________________
00:00 – Introduction to the episode and our guest
02:29 – The evolving role of AI in work
04:20 – Real-world applications of AI in clinical reporting
06:24 – Measuring success: KPIs and user feedback
08:36 – Acceleration in clinical development with AI
10:20 – Challenges in implementing AI tools
14:49 – Collaboration in AI development
19:28 – Future developments and initiatives at Roche
24:02 – Redefining processes with AI
29:06 – Automation vs. AI: Finding the right approach
32:17 – An unsolved problem in the field
35:00 – Outro and closing remarks
#datascience #genAI #clinicaltrials #clinicalreporting #AIinpharma #GenerativeAI #Roche #AIautomation #pharmaceuticalindustry #AIchallenges #AIbot #chatbots #copilot #codingassistant
In this episode, we speak with Vincent Shen, Senior Principal Data Scientist at Roche, about the actual impact of generative AI in clinical data analysis. Vincent shares insights from his work implementing AI-powered tools, such as knowledge chatbots and coding assistants, that support clinical reporting at Roche. We discuss the evolving role of AI, the challenges of integrating it into clinical trial workflows, and the importance of rethinking the processes to ensure effective AI use.
In this episode::
- AI will augment rather than replace data scientists in clinical trial workflows.
- Roche has successfully deployed a GenAI-powered knowledge bot, saving an estimated 25 hours per 100 questions.
- Coding assistants help data science teams accelerate time-to-insight.
- Providing detailed specifications is essential for AI to deliver accurate and meaningful results.
- Training AI models requires extra effort to ensure they correctly interpret context and domain-specific requirements.
- The next big step is to rethink how processes are designed, not just automate existing ones.
Vincent Shen
linkedin.com/in/vincent-shen-vs
Nat Chrzanowska
linkedin.com/in/nat-chrzanowska
__________________________________________
More about Appsilon: ► appsilon.com
Appsilon empowers pharmaceutical and life sciences companies to leverage open-source technology for faster, data-driven decision-making in regulated environments. Schedule a free consultation with our expert
► appsilon.com/contact-us
We design scalable and user-friendly Shiny dashboards to help you make data-driven decisions.
► appsilon.com/services/data-dashboards
We design and implement Statistical Computing Environment for R and Python for efficient data analysis:
► appsilon.com/services/sce
We design, implement, and optimise data analysis environments so you can focus focus on insights and innovation.
► appsilon.com/services/platform
Where Technology Meets Science podcast is available on all podcasting platforms:
► Spotify: https://hubs.li/Q037KddD0
► Apple Podcasts: https://hubs.li/Q037KlLl0
► YouTube: https://hubs.li/Q037Kpth0
__________________________________________
For more insights about how technology helps scientists push the boundaries of data analysis and reporting check out our blog:
► appsilon.com/blog
LinkedIn: linkedin.com/company/appsilon
__________________________________________
00:00 – Introduction to the episode and our guest
02:29 – The evolving role of AI in work
04:20 – Real-world applications of AI in clinical reporting
06:24 – Measuring success: KPIs and user feedback
08:36 – Acceleration in clinical development with AI
10:20 – Challenges in implementing AI tools
14:49 – Collaboration in AI development
19:28 – Future developments and initiatives at Roche
24:02 – Redefining processes with AI
29:06 – Automation vs. AI: Finding the right approach
32:17 – An unsolved problem in the field
35:00 – Outro and closing remarks
