6 October 2025 · 1 min read
An interesting talk by Roche's Pierre Fischer at BioTechX about their Data Mesh Operating model.
I attended an interesting talk by Roche's Pierre Fischer at BioTechX about their Data Mesh Operating model, covering domain ownership, a dedicated data marketplace, and how their platform reduced dashboard delivery from months to under a week.
Filed under AI in Pharma
An interesting talk by Roche's Pierre Fischer at BioTechX about their Data Mesh Operating model.
Some key insights:
-
Domain ownership is crucial, with proper govnernance and oversight, and dedicated self sufficient product teams
-
A dedicated data market place so that teams can find the data
-
Simplified data access policies so people can actually get the data, with a self service data infrastructure
Roche's data operations platform has been a great success, turning data truly into a product, helping teams accelerate. In one example they were able to reduce time to a new dashboard from months to less than a week.
The data platform includes the ability of securely sharing data with external parties.
Pierre highlighted some interesting challenges they encountered, eg moving over more responsibilities from IT to the actual business users, as well as the challenges of getting to industrial grade robustness for the Snowflake-based platform.
Related insights
17 Mar 2026
Roche’s hybrid-cloud AI factory: why adoption matters more than GPUs
3 Mar 2026
Lab-in-the-loop drug discovery is an operating model problem, not a model problem
26 Feb 2026
Clarivate LS&H sale: what GenAI means for pharma intelligence businesses
17 Feb 2026
Big Pharma AI deals: why binding intelligence matters more than “AI”