Claws and Agents in Production. Now How to Scale.
So you built the agent.
It worked beautifully in the demo. Your boss was impressed. Your investor nodded. You felt like a genius for eleven minutes.
Then production happened.
The voice agent started stuttering a when call volume spiked. The coding agent hallucinated a function that doesn't exist in any language ever invented by humans. The enterprise workflow that took 30 seconds in testing now takes 3 minutes and sometimes just sits there, thinking, like your uncle after two whiskeys.
You are not alone. Everyone building agents in India right now is hitting the same wall. They're just not talking about it because nobody wants to be the one who says "my agent breaks."
We are getting 15-20 of these people in one room. On a Friday afternoon. In Bangalore. With food.
First half, you put your biggest problem on a board. Coding agents, voice agents, enterprise things that were supposed to be intelligent. We find who broke what and who fixed what.
Second half, we go underneath. Inference. Where the money actually goes. Where the latency actually hides. The bit nobody looks at until the bill arrives.
Vamshi Ambati and I (Rajan) facilitate this. We co-wrote the book on this. Literally. It's called Peak Inference.
Friday, March 27, 2026 2:30 - 5:00 PM IST Bangalore. We'll tell you where once you're in.
Chatham House Rule. Which means you can talk about what was said but not who said it. So you can actually be honest.
To get in, tell us:
What agents are you running in production?
How many runs a day? Loops per run?
The one problem that's making you quietly lose your mind
No recordings. No slides. No LinkedIn posts with "5 key takeaways."
Just people who build agents, in a room, eating food, comparing scars.