Pranav Zagade

About

Built and shipped an LLM-powered support agent for Community Dreams Foundation that automated intake and Jira ticket creation using RAG, structured outputs, and strong production guardrails. Demonstrated practical production AI experience with 92% routing accuracy, 98% uptime, and a 31% improvement in first-response accuracy, plus hands-on work in observability, evals, idempotency, and failure handling.

Key Strengths

• Shipped LLM support agent to production
• Automated support intake, classification, and Jira ticket routing
• 92% routing accuracy
• 98% uptime in production
• Improved first-response accuracy by 31%
• Strong guardrails for predictable agent behavior
• Designed multi-step agent workflows with tool calling and structured outputs
• Built evaluation, monitoring, and feedback loops for production agents
• Solved duplicate ticket creation with idempotency and deduplication
• Handled messy operational data with validation layers and safe fallbacks
• Refactored fragile SQL-based analytics pipeline into staged, validated workflows to improve reliability and debuggability
• Uses incremental rollout strategies (parallel pipeline runs, result comparisons) to manage migration risk and preserve data integrity
• API reliability practices in FastAPI (Pydantic validation, clear separation of layers, explicit error handling/status codes)
• Security implementation experience (JWT auth, role-based permissions, row-level security)
• Identifies and mitigates silent data pipeline failure modes via validation gates to prevent bad data propagation
• Reduced backend module size ~80% through incremental refactoring, improving maintainability and reliability

Experience

Education

Awards

• Dean’s List Holder (6x) - Arizona State University
• Magna Cum Laude - Arizona State University

Skills

Languages