
How DoorDash knew its AI agent got worse
DoorDash had a production agent running on Claude Sonnet 4.6.
Everything I know about building software, I learned from people who shared it openly. This is my turn to share back. I'm Zahiruddin Tavargere (Zahere), and I write about AI agents, LLMs and engineering, with practical lessons you can use.

DoorDash had a production agent running on Claude Sonnet 4.6.
Three places your retrieval pipeline breaks, and what I measured when I fixed them.
Open any AI engineer or Forward Deployed Engineer job post and you’ll see some version of the same line: “RAG experience on real enterprise data.”

I didn’t realize I was doing Forward Deployed Engineering until someone else gave it a name.
I didn’t write my first real line of code until my early 30s
For years, we’ve assumed that production AI systems need to call a hosted model every time they need to make a decision.
What if the next breakthrough in AI isn’t another chatbot...

If you’ve built an MCP client with dozens—or hundreds—of tools, you’ve probably run into the same question:
I spent more time on one mistake than I expected.