Super Data Science: ML & AI Podcast with Jon Krohn 1023: Agentic AI Skills That Matter Now, with Aishwarya Srinivasan
28 snips
Sep 1, 2026 
Aishwarya Srinivasan, an AI practitioner and educator who has worked at Google, Microsoft, and IBM, explores what really matters in the age of agentic AI. She breaks down why engineering judgment beats vibe coding, how whole-system evaluations expose real-world risks, and why reinforcement learning is back. She also shares the MIND framework, loop engineering, and the career advantage of relentless experimentation.
AI Snips
Chapters
Books
Transcript
Episode notes
600 Mentors Turned Informal Help Into Gen Academy
- Aishwarya Srinivasan began Illuminate AI during the pandemic after LinkedIn mentoring requests became unmanageable.
- Its first cohort received 600 mentor applications, prompting her to build structure and eventually expand into Gen Academy.
Cheap Code Makes Judgment The New Bottleneck
- Cheap AI-generated code lowers the barrier to building demos but does not remove architecture, security, scalability, or product judgment.
- A mindless ten-minute prototype creates no moat because competitors can reproduce it just as quickly.
Agent Evaluation Must Test The Entire System
- Agent uncertainty compounds across model decisions, tool calls, loops, harnesses, permissions, databases, and external failures.
- Production evaluation must test the complete path from user input through tools, failures, security threats, and final response.
