Description
Key Responsibilities
- Own the end-to-end technical architecture of the agentic platform — including agent orchestration framework, HITL engine, exception routing, integration layer, intelligence services, canonical data model, and workflow state store — ensuring all components are cohesive, scalable, and multi-tenant by design.
- Define the multi-agent design patterns used across the platform — agent boundaries, tool use contracts, inter-agent communication protocols, confidence thresholds, fallback routing, and human handoff triggers — and govern their consistent application across all workflow implementations.
- Lead the design of the workflow state management architecture — covering execution context persistence, checkpointing, idempotent retries, long-running process patterns, and safe resume — ensuring finance workflows that span days or weeks behave correctly under failure conditions.
- Define the platform's multi-tenancy architecture — tenant isolation, per-client configuration, shared service design,



and data segregation — so that a single platform deployment can serve multiple clients safely and efficiently.
- Establish the observability and governance architecture — structured logging, distributed tracing, model performance monitoring, SLA tracking, audit trail design, and compliance logging — ensuring the platform meets the control requirements of regulated finance environments.
- Evaluate and select foundational technologies — LLM providers, agentic frameworks (LangGraph, AutoGen, CrewAI), orchestration engines (AWS Step Functions, Temporal), vector databases, and messaging infrastructure — with transparent justification for each choice.
- Lead technical design reviews and architecture governance — reviewing critical implementation decisions made by AI engineers, backend engineers, and integration engineers, and ensuring they align with the platform's intended design.
- Work directly with client technical stakeholders during pre-sales, solutioning

📌 Agentic Architect (India)
🏢 EXL
📍 India