sasanka voleti

About

Built and deployed a production AI/ML SRE copilot that uses RAG over real-time Splunk signals plus deployment/runbook data to generate grounded incident summaries and next steps, cutting time-to-contact by 30%. Treats the knowledge corpus like a production dataset (quality gates, semantic chunking, metadata enrichment) and runs golden-dataset automated evals to ensure reliability, while partnering closely with ops/support leaders through discovery sessions and metric-driven demos.

Experience

Education

Key Strengths

• Built and deployed an AI/ML SRE incident-triage copilot to production
• Delivered 30% reduction in time to contact for incidents
• Designed grounded, structured incident summaries using RAG with real-time operational signals
• Production-grade knowledge ingestion with quality gates, semantic chunking, deduping, normalization, and metadata enrichment
• Reliability-focused evaluation approach using golden datasets and automated evals on every change
• Pragmatic model selection: moved from large LLM + pure vector retrieval to smaller model + hybrid retrieval with strict metadata filters based on offline eval results
• Effective collaboration with non-technical ops and support stakeholders via discovery sessions, measurable goals, and lightweight demos/metrics