The TWIML AI Podcast (formerly This Week in Machine Learning & Artificial Intelligence) ML Platforms for Global Scale at Prosus with Paul van der Boor - #468 [TWIMLcon Sponsor Series]
Mar 29, 2021
Join Paul van der Boor, Senior Director of Data Science at Prosus, as he shares his journey from aerospace engineering to leading data science at a major tech firm. He discusses the hurdles AI builders face transitioning from demos to real-world applications and the importance of thorough evaluations. Paul explores building ML capabilities across diverse teams in a global organization, emphasizing collaboration and standardization. He also reflects on insights gained from a recent industry conference, highlighting networking as a key to overcoming challenges.
AI Snips
Chapters
Transcript
Episode notes
Prosus's Diverse AI Challenges
- Prosus, a large consumer internet company, faces diverse AI challenges across its portfolio.
- These range from food delivery logistics to content moderation on secondhand marketplaces.
No One-Size-Fits-All ML Platform
- A one-size-fits-all approach to ML platforms isn't suitable for diverse business needs.
- Real-time use cases, like food delivery, have different requirements than batch processing, like content moderation.
Challenges and Benefits of Centralized Platforms
- Building a common ML platform is difficult due to varying business needs and path dependency on existing infrastructure.
- Sharing knowledge about tooling and adopting reusable "recipe books" for platform components can be beneficial.