Saratchandra Patnaik

Incident: A critical client's live stream fell back to rescue content due to an audio silence condition — a silent failure that required tracing logs across EC2 instances, Kubernetes pods, and multiple microservices to diagnose.

Finding: The provider had primary and secondary input streams but no automated health check or switching mechanism. Audio silence propagated undetected until the playout system gave up and triggered rescue content.

Fix: Implemented threshold-based failover logic that monitors audio presence and switches to the healthy secondary stream when silence persists beyond a defined window — eliminating rescue fallback and hardening the client's stream against input failures.

Scope: Amagi was migrating from a legacy Terraform-based provisioning model to a new GitOps-style deployment environment where YAML configs committed to Git triggered ArgoCD CI/CD pipelines to provision and manage resources. There was no established end-to-end runbook yet — I became the first engineer to deploy a real customer environment in the new model.

Finding: The migration wasn't just a tooling change — it affected backend service rollout, microservice deployment, update handling, and rollback behavior across the platform. Applied the new model to DraftKings, a complex customer with substantial customization, as the real-world production validation.

Fix: Mapped the full new deployment workflow end to end, validated rollout safety and rollback paths via Git-versioned state, and successfully delivered the DraftKings environment in ~1 week — down from the prior typical timeline of ~4 weeks. Standardized the process into a repeatable organizational runbook so future deployments wouldn't depend on tribal knowledge.

Scope: Investigated manifest behavior, proxy chain issues, and SCTE marker handling across production streaming services — crossing platform, infrastructure, and client-facing boundaries to isolate live traffic failures.

Codec & Protocol Work: Debugged H.264/H.265 codec configurations and AWS MediaLive settings to resolve client-facing playback errors across WebRTC and WebRTS workflows — reduced playback errors by 95%.

Backend: Built non-blocking FastAPI services for high-concurrency streaming control; integrated StreamDeck hardware for real-time broadcast operator control.

Scope: Amagi built native subtitle generation to eliminate dependency on customers supplying external .srt files. The feature had never been deployed for a real customer. I led the first-ever production rollout across both channel types.

Finding: Two distinct workflows existed — Capsequo (real-time speech-to-text via Google API) for Live channels like sports/news, and Akashvani (serverless, S3-triggered) for FAST channels. No proven deployment path existed end to end.

Fix: Mapped deployment requirements for both workflows, provisioned required components during channel setup, validated feature flags and service dependencies, and ran end-to-end testing through the full subtitle lifecycle. After successful rollout, wrote a full runbook and trained the wider engineering org — converting a first-time deployment into a repeatable organizational standard.

GitOps: Moved deployments to ArgoCD for version-controlled, reproducible releases — 99.9% rollout stability and 30% higher deployment frequency across 15+ services.

IaC: Automated customer environment provisioning with Terraform (VPCs, subnets, CDN) — cut manual setup by 90% and CDN delivery failures by 95%.