Talks

Debugging AI agent failures at scale is hard

Agents fail differently, and sampling hides it. Keeping every span from 100,000 calls a day in production at Fello — and debugging them from anywhere over MCP.

Akhilesh Krishnan · Founding Engineer

Building a logs engine faster and cheaper than ClickHouse

Observability is a data engineering problem. Object storage plus a serverless architecture as the building blocks for a logs engine that is both faster and cheaper.

Handling large tool outputs in agentic workflows

Speeding up Next.js production builds with OpenTelemetry

Dashboard-less debugging with Oodle AI

Gaurav Maheshwari · India Site Technical Lead

Can you have cake and eat it too? Fast and cheap observability

Why observability became expensive, complicated and slow, and how legacy and state-of-the-art approaches trade cost against query performance.

Kiran Gollu · Co-founder & CEO

A new approach to observability: cut costs and complexity with Serverless and S3

The typical metrics observability stack, where its cost and complexity come from, and why serverless and S3 make now the right time to rethink it.

Kiran Gollu · Co-founder & CEO

Serverless Prometheus

Tao Wan · Staff Engineer

Architecting Prometheus for cost-efficiency

Kiran Gollu · Co-founder & CEO