Challenge
Make CI failure diagnosis structured enough that agents can reason over state, retry safely, and expose each step to the developer.
Back to work
AI Infrastructure
Autonomous CI/CD Healing Agent - RIFT
Multi-agent CI/CD debugging and healing system.
A CI/CD debugging system that streams failure analysis, executes sandboxed tests, and proposes reviewable LLM-generated patches.
LangGraphNode.jsReactDockerGroq
Agent Loop
LangGraph coordinates bounded retries, real-time progress streaming, sandboxed execution, and reviewable patch output.
- Up to 5 retry iterations.
- SSE streaming for debugging progress.
- Docker-sandboxed pytest execution.
Outcome
The project reached AIML Track Finalist status at RIFT 2026 and demonstrates agentic infrastructure with guardrails.