Kafka at EchoStar: Less Ops Overhead, More Guardrails

Akshara Madduluri and Shreshth Rathod are data engineers at EchoStar. Akshara owns the observability stack and Kafka and is now moving into hosting AI gateways; Shreshth works across Python, Spark, Kafka and Grafana. Recorded at Current, they talk through what running Kafka costs a small team day to day, and what changed when they stopped managing brokers themselves.

In this conversation

• →Why scaling a self-managed Confluent Platform cluster on EC2 took eight to ten hours of Ansible runs
• →Moving to a diskless, S3-backed Kafka platform on EKS that autoscales under load
• →The one feature on their Kafka wishlist: rebalancing without the inter-AZ bill
• →What it takes to trust an agent in production: scoped skills and explicit guardrails