About this role
AI/ML Engineer (Agentic Systems) Location: Pune Experience: 1-2 years Employment type: Full-time
About the Role We are building an end-to-end agentic AI system that plans, uses tools, and completes multi- step tasks reliably. As an AI/ML Engineer, you will help build and improve the agent itself, working closely with senior engineers who will guide your design decisions. You'll own well- scoped pieces of the system and grow into broader ownership over time.
Responsibilities ● Build and extend LLM-powered agents, including tool/function calling, memory, and multi-step workflows ● Implement agent workflows using frameworks such as LangGraph, LlamaIndex, or similar ● Write, test, and refine prompts, system instructions, and tool schemas ● Help build RAG pipelines (chunking, embeddings, retrieval) alongside data engineers ● Run experiments comparing models and prompts on quality, latency, and cost ● Create and maintain evaluation datasets and automated tests to track agent quality ● Implement guardrails such as output validation, fallbacks, and human-in-the-loop checkpoints ● Add tracing and logging so agent failures can be debugged step by step ● Work with backend engineers to connect the agent to internal systems and APIs ● Document experiments, findings, and agent behavior clearly for the team
Required Qualifications ● Bachelors or Master’s degree in Computer Science, AI/ML, or a related field (or equivalent practical experience) ● 1-2 years of experience in software engineering, ML, or applied AI work ● Solid Python skills, including Git, testing, and writing clean, readable code ● Hands-on experience building with LLM APIs (OpenAI, Anthropic, open-source models, etc.) ● Working knowledge of prompt engineering, structured outputs, and tool/function calling ● Basic understanding of RAG and vector databases (e.g., pgvector, Pinecone, FAISS) ● Familiarity with REST APIs and basic cloud concepts
Preferred Qualifications ● A portfolio project, open-source contribution, or internship involving an agent or LLM application ● Exposure to LLM observability tools (LangSmith, Langfuse, or similar) ● Familiarity with Docker and basic CI/CD ● Awareness of LLM risks such as prompt injection and hallucination ● Interest in or exposure to Model Context Protocol (MCP)