Sagar Shankaran

About

Built and deployed a production real-time voice AI support agent that answers inbound calls, identifies callers, troubleshoots via a knowledge base, and automatically creates/updates tickets with escalation to humans when needed. Demonstrates strong reliability/latency engineering (streaming, schema validation, idempotency, DB constraints) and uses LangGraph state machines plus OpenAI Agents SDK for multi-agent routing, with KPI-driven testing and monitoring.

Experience

Education

Key Strengths

• Built and deployed a real-time voice AI support agent that identifies callers, resolves issues, and creates/updates tickets with human escalation
• Latency optimization for real-time voice interactions via end-to-end streaming and background tool execution
• Reliability engineering for tool-calling under noisy voice inputs using strict schemas and validation layers
• Handled duplicate tickets and race conditions using idempotency keys, DB uniqueness constraints, and upsert workflows
• Designed agents as guarded state machines for deterministic flow, safer retries, and easier debugging (LangGraph)
• Implemented multi-agent routing with triage-to-specialist handoffs for improved control and reliability (OpenAI Agents SDK)
• Strong evaluation discipline: KPI-driven metrics, scenario libraries from production logs, golden-set offline evals, and live monitoring/alerts
• Effective collaboration with non-technical stakeholders via iterative demos and requirement clarification