Machine Learning Engineer

Overview

Design, build, and operate production AI agents and tool using Agentic AI frameworks in Python. Own agentic workflows end to end across planning, reasoning, tool calling, retrieval, evaluation, security, and observability.

Responsibilities

• Implement agentic workflows in Python using an agent framework. Semantic Kernel is preferred. Experience with LangGraph, AutoGen, CrewAI, or equivalent is valuable.
• Build reliable tool and function calling flows with planning, memory, and conversation orchestration.
• Develop retrieval pipelines for RAG and hybrid search including ingestion, chunking, embeddings, ranking, query planning, grounding, and caching.
• Ship agents to production on cloud platforms. Integrate identity, networking, cost controls, and runtime observability.
• Instrument applications with tracing, metrics, and logs.
• Establish evaluation workflows using prompt and flow tests for offline, batch, and A/B scenarios.
• Collaborate with product, data, and security to harden systems using rate limits, retries, timeouts, and circuit breakers. Provide documentation and mentorship.

Must have

• 4 to 5 years hands on GenAI application experience before moving to agentic work.
• Strong Python 3.11 plus skills including typing, asyncio, packaging, testing with pytest, profiling, and CI/CD.
• Production experience with at least one agent framework. Semantic Kernel preferred. LangGraph, AutoGen, or CrewAI acceptable.
• Expertise in tool and function calling, schema design, argument validation, and multi-step planning.
• Retrieval systems experience with vector stores and hybrid search. Familiarity with grounding strategies and retrieval evaluation.
• Cloud deployment experience with containers, secrets and identity, networking, monitoring, and alerting.
• Observability and evaluation proficiency including tracing, metrics, log aggregation, experiment design, and promotion criteria.
• Security and safety fundamentals including prompt injection defenses, content policy enforcement, tool sandboxing, and PII handling.
• Clear technical communication and collaborative code reviews.

Good to have

• Multi agent patterns such as task decomposition, coordinator worker, and human in the loop.
• Deeper Azure experience including Azure AI Search and related AI platform services.
• Evaluation depth with regression suites, red teaming, and data driven guardrails.
• Search and data store depth with Elasticsearch, Pinecone, pgvector, or Azure AI Search.
• Frontend integration for agent UIs with streaming and tool traces. Secure API design for tools and connectors.
• DevOps skills with Docker, Kubernetes, GitHub Actions or Azure DevOps, infrastructure as code, and secrets management.