Reliability as code
Infra Reliability Orchestrator
An evidence-backed system that scores service reliability, drafts remediation PRs, and prevents safeguards from quietly disappearing.
Problem
Reliability reviews were fragmented across manifests, monitoring configuration, team knowledge, and lessons buried in historical incident reports. That made gaps difficult to spot and safeguards easy to remove unintentionally.
Approach
Built an orchestrator that gathers evidence from Kubernetes resources, alerting rules, and more than 100 historical RCAs. It evaluates each service across 90 reliability checks, explains the evidence behind the score, and turns actionable gaps into remediation pull requests.
Contribution
Designed the evidence model and orchestration flow during Razorpay's Day 0 hackathon, connected reliability findings to code changes, and added a pull-request gate that catches changes removing existing safeguards.
Measured outcome
Created an auditable reliability score for every service, auto-drafted more than 70 remediation pull requests, and shifted incident learning from passive documentation into preventive engineering.