E163: Using Feedback Loops to Optimize LLM-Based Applications

Viraj Mehta is the Co-Founder & CTO of TensorZero which is an open-source infrastructure platform that creates a feedback loop for optimizing LLM applications. Their open source project helps users turn production data into smarter, faster, and cheaper models.

In this episode, we dig into:

• The benefits of feedback loops for LLMs

• Helping their users choose the best underlying models for their applications

• "Recipes" as a potential monetization path

• The most common optimizations for LLM-based apps

• Educating users on what's possible with LLMs