Super Data Science: ML & AI Podcast with Jon Krohn 971: 90% of The World’s Data is Private; Lin Qiao’s Fireworks AI is Unlocking It
50 snips
Mar 3, 2026 
Lin Qiao, CEO and co-founder of Fireworks AI and former Meta engineering leader, builds platforms that unlock private enterprise data for custom open-source models. She discusses autonomous intelligence, coding agents that act like junior engineers, why enterprises should bet on open models, and how Fireworks optimizes model, hardware, and evaluation choices for real-world use.
AI Snips
Chapters
Transcript
Episode notes
Private Enterprise Data Holds Most Intelligence
- Over 90% of the world's valuable intelligence sits in private enterprise apps and never enters foundation models.
- Lin Qiao argues activating that private data via continuous automated customization yields millions of specialized models tailored to applications.
Autonomous Intelligence Replaces Expert Bottlenecks
- Autonomous intelligence means continuously and automatically customizing models and inference for each application rather than relying on a single AGI model.
- Lin says this automation should make model-product co-design accessible to many developers without deep specialist burdens.
Specialize Models With Reinforcement Style Tuning
- Use reinforcement-style tuning to specialize models for a product by defining positive and negative feedback and iterating.
- Lin compares this to selective breeding and recommends continuous specialization to create a private moat for apps.