Internal tool concept for L1/L2 teams to reduce manual stack-trace analysis and accelerate RCA.
Identified gap from 300+ manual RCA cycles; built as internal tooling proof-of-concept to address a real support bottleneck.
•Problem: Support teams spending excessive time manually analyzing Java stack traces and error patterns across multiple systems for enterprise customers.
•Solution: Built automated portal with file uploads, API integrations, and Groq API integration to parse errors, targeting 40% deflection in L1 tickets and faster MTTR for L2.
•Architecture: MERN stack with React frontend for file uploads and real-time diagnostics, Express/Node backend orchestrating LLM calls via Groq API, and MongoDB for log storage and pattern history.
•Impact: Reduced average RCA time from 45 minutes to under 15 minutes per incident, enabling L1 teams to self-serve on common error patterns without escalation.
Stack: React, Node.js, Express, MongoDB, Groq API, REST APIs
