Senior / Lead - AI/ML Engineers - Aziro

AI Agentic Platform Engineer with LangChain,LangGraph,LangSmith

We’re seeking seasoned engineers to build production-grade multi-agent AI platforms and capabilities. This role emphasizes deep hands-on experience with LangChain, LangGraph, and LangSmith, agent-to-agent orchestration, and foundational AI platform architecture with integrated intelligence layers and AWS services (including AWS Bedrock and Agentic AI).

What you’ll do

• Implement, and operate multi-agent AI workflows using LangChain and LangGraph (stateful graphs, event-driven nodes, tool routing, memory, retries, and guardrails).
• Design and build foundational AI platform capabilities with integrated intelligence layers (reasoning, planning, routing, policy enforcement) that can be reused across products and teams.
• Define and implement agent-to-agent communication protocols, design robust AI personas, and create task flows that go beyond simple query–response (e.g., long-running tasks, collaborative agents, human-in-the-loop workflows).
• Build complex multi-agent systems: agent-to-agent collaboration, hierarchical planners/executors, and tool-augmented reasoning (ReAct, Reflexion, Tree-of-Thoughts where appropriate).
• Use LangSmith for tracing, observability, debugging, and evaluation of agentic workflows; feed insights back into platform design and optimization.
• Integrate AWS services (including Agentic AI and AWS Bedrock) into the platform for model access, orchestration, and secure deployment patterns.
• Design reusable platform primitives: prompt/graph registries, tool catalogs, provider routing/fallbacks, secrets/key management, policy/guardrail layers, and evaluation harnesses.
• Expose agents and workflows via well-versioned APIs/SDKs; partner with product/platform teams to standardize patterns and publish internal best practices.
• Integrate and optimize retrieval where needed (vector stores, chunking, hybrid search), with a strong bias toward agentic solutions for complex, multi-step tasks.
• Own production readiness for multi-agent systems: cost/latency optimization, caching/semantic caching, observability, tracing, rate limiting, and incident response.
• Benchmark across models/providers (OpenAI, Anthropic, AWS Bedrock, Azure/OpenAI, local/vLLM/TGI) and tune for reliability, safety, and task success.
• Mentor engineers, lead design/architecture reviews, and contribute to the internal GenAI platform roadmap, especially around multi-agent and platform primitives.

Must-have skills

• 7+ years of software engineering (Python preferred), with 2+ years building LLM-powered Multi Agent Applications in production.
• Practical experience with LangChain, LangGraph, and LangSmith for building and operating AI platforms:
• Chains, agents, tools, retrievers, routers, and memory stores.
• Designing REST/gRPC endpoints and/or SDKs for agents, tools, and workflows; authn/authz; schema and backward compatibility.
• LangGraph state machines, nodes/edges, interrupts, checkpointers, event streams, and retries.
• LangSmith for tracing, observability, debugging, and evaluating agentic workflows.
• Designing multi-agent architectures, agent-to-agent communication protocols, role-specific AI personas, and task flows beyond traditional query–response.
• Observability and tracing for LLM systems (LangSmith, OpenTelemetry; metrics, logs, spans).
• Familiarity with AWS services including Agentic AI and AWS Bedrock for model orchestration, deployment, and governance.
• Cloud-native deployment with Docker and Kubernetes on AWS /Azure; CI/CD.
• Solid engineering practices: testing strategies for LLM apps (unit, contract, eval tests), performance profiling, and cost governance.

Good to have

• Experience with additional agent frameworks (CrewAI, AutoGen, LlamaIndex agents) and OpenAI Assistants.
• Retrieval and data tech: vector stores (Pinecone, Weaviate, FAISS, Milvus), hybrid/BM25, knowledge graphs.
• JavaScript/TypeScript exposure for SDKs or front-end integrations.