Description

You will design, build, and productionize specialized agent workflows for enterprise systems, owning the end-to-end agent lifecycle from prompting to deployment.

This role is on-site.

Responsibilities

• Implement agents using Strands Agents SDK (or equivalent) with integrations across NuOrder, Freshservice, Snowflake, and GXS.
• Design prompts, workflows, and state machines for planning, retries, and fallbacks, including hallucination detection and HITL escalation.
• Build and operate RAG pipelines, embeddings, vector search, and retrieval strategies to encode business rules and PO creation logic.
• Develop REST APIs and lightweight SDK wrappers (Python/TypeScript) with authentication, retries, typed schemas, and OpenTelemetry context propagation.
• Monitor production-grade dashboards using Grafana/Prometheus/APM and integrate evaluation tools like LangSmith, Langfuse, or DeepEval.

Required Skills

• Advanced Python with 5+ years of experience.
• Hands-on LLM/agent development using AWS Bedrock, LangChain/LangGraph, Strands, AutoGen, or CrewAI.
• Expertise in advanced prompt engineering: few-shot prompting, ReAct patterns, structured prompting, and guardrails.
• RAG expertise including embeddings, vector search, and experience with vector DBs like pgvector, Pinecone, or Weaviate.
• Production integration skills: REST/GraphQL APIs, message queues (SQS/Kafka/RabbitMQ), and strong error handling.
• Platform hygiene: Microservices (FastAPI), Docker, Kubernetes, Git, CI/CD, and cloud fluency (AWS/Azure/GCP).
• Observability: Grafana, Prometheus, APM dashboards, OTEL traces, and SLO-based alerting.

Preferred Skills

• Experience with AutoGen, CrewAI, LlamaIndex, graph-based planners, or toolformer patterns.
• Vector DBs and retrieval: Milvus, Weaviate, FAISS, or hybrid search strategies.
• LLM optimization: Context caching, quantization, memory management, and prompt A/B testing.

Key Skills

Education

Any Graduate