AI/ML Engineer - Agentic AI

ROLE : AI / ML Engineer- Agentic AI

EXPERIENCE : 48 years

LOCATION : Pune, India (Onsite)

EMPLOYMENT TYPE : Full-time, Permanent

OPENINGS : 1

AVAILABILITY : Immediate joiner preferred

About The Role

Netscribes is building a dedicated engineering pod for a large enterprise client engagement focused on product support, data engineering, and AI-driven automation. As part of this team, the AI / ML Engineer will own the design and delivery of agentic AI workflows that automate knowledge work, surface insights, and augment existing product capabilities across multiple business units. The role calls for hands-on engineering not prototyping. You will build production-grade multi-agent systems using LangGraph, CrewAI, and Databricks Agentbricks, integrated into an Azure hosted product stack. You will work alongside data engineers, full-stack developers, and a project lead, shipping iteratively within an hours-driven delivery model.

Key Responsibilities

• Design and build multi-agent pipelines using LangGraph and CrewAI; define agent roles, memory strategies, and tool-use patterns for each workflow.
• Integrate LangSmith for end-to-end tracing, prompt versioning, evaluation harnesses, and production observability across all agent deployments.
• Develop and deploy agentic solutions on Databricks using Agentbricks; manage MLflow experiments, Unity Catalog assets, and model serving endpoints.
• Collaborate with data engineers to wire agent pipelines to existing Python and Databricks data pipelines; ensure data freshness and schema compatibility.
• Instrument and monitor deployed agents latency, token spend, failure modes and iterate based on real usage.
• Write clean, testable Python; document agent graphs, prompt libraries, and tool specifications for the wider pod.
• Contribute to the knowledge management system scoped into the engagement (setup, ingestion pipelines, retrieval tuning).
• Participate in sprint reviews, demo working agents to client stakeholders via the project lead, and incorporate feedback rapidly.
  
  

Required Skills & Experience

• 4-8 years of software engineering experience, with at least 2 years focused on LLM application development or agentic AI systems.
• Hands-on production experience with LangGraph state machines, conditional routing, human-in-the-loop checkpoints.
• Working knowledge of LangSmith tracing runs, building evaluation datasets, monitoring prompt drift.
• Practical experience with CrewAI defining agents, tasks, and crew orchestration for multi-agent scenarios.
• Experience on Databricks, including Agentbricks for agent deployment, MLflow for experiment tracking, and Delta Lake for data access.
• Strong Python skills : async patterns, tool/function calling, API integration, prompt templating.
• Familiarity with Azure OpenAI or equivalent LLM provider APIs (rate limits, context window management, cost control).
• Ability to read and contribute to broader product codebases (NodeJS, Java, React) for integration touchpoints.
• Strong written communication agent graphs and prompt libraries must be documented clearly for handoff.
  
  

Nice To Have

• Experience with RAG architectures chunking, embedding models, vector stores (Azure AI Search, Chroma, FAISS).
• Familiarity with AutoGen or other agentic frameworks beyond LangGraph and CrewAI.
• Exposure to Streamlit for lightweight internal tooling or agent UIs.
• Prior work in an hours-driven or sprint-based managed services model.
• Azure AI Studio or Azure ML workspace experience.
  
  

Education

• B.E. / B.Tech. / M.Tech. in Computer Science, Electronics, or a related engineering discipline.
• Relevant certifications (Databricks Generative AI Engineer Associate, Azure AI Engineer Associate) are a plus.
  
  

What Success Looks Like

• Within 60 days : at least one agentic workflow is live in a production or staging environment, instrumented in LangSmith, and delivering measurable automation.
• By month 4 : the knowledge management system is operational, with ingestion pipelines running and retrieval quality validated against real queries.
• By end of pilot : agent layer is stable, documented, and extensible the project lead can scope new agent tasks in hours without senior engineering involvement.
  
  

(ref:hirist.tech)