Talks I offer · each one has an essay behind it
Talks
I build and measure GenAI products inside a large enterprise, and run small AI systems of my own. These are the talks I can give from that work. Read the essay first to see the argument.
Topics
- 01
Your AI Metric Is Meaningless Without a Benchmark
The same number reads as great, fine or not enough depending on the yardstick. How to source one, and how to measure an AI product without an ML team.
Read first: The benchmark essay · The $22-a-month sentiment pipeline
- 02
Decision-Time Logs Beat Replay: Proving an AI Fix Worked
A support bot kept telling people to go away. Finding the bug was easy; proving the fix nearly fooled me. What to log, and why replaying old traffic misleads.
Read first: My AI bot kept telling people to go away
- 03
A $0 Memory Every AI Tool Can Read
How a Mac mini, SQLite and MCP give Claude, Cursor and ChatGPT one shared memory, and what that suggests for enterprise AI products that keep forgetting their users.
Read first: I built a second brain that talks back
Formats
- Conference talk, 20–40 min
- Podcast or fireside chat
- Team workshop
- Panel
Bio
Aashish Jagini is a senior product manager at McKinsey, where he shipped the firm's first production GenAI support agent. He started as a software engineer and still builds the AI tools he uses.
To book
Email imajhere@gmail.com with the event, audience and date. Based in Frisco, TX; happy to travel or join remotely.