IN_Senior Associate_Agentic AI Developer using Python_GCC_Advisory_Bangalore

Responsibilities: Agent Design & Orchestration . Design and develop autonomous and semi-autonomous AI agents capable of multi-step reasoning, planning, and decision-making . Build multi-agent systems with defined roles, inter-agent communication, and collaborative task execution . Implement agent orchestration frameworks (LangGraph, AutoGen, CrewAI, or custom-built) for workflow coordination . Define and manage agent memory layers - short-term (context window), long-term (vector stores), and episodic memory . Develop tool-use and function-calling pipelines enabling agents to interact with APIs, databases, and external services LLM Integration & Prompt Engineering . Integrate and fine-tune interactions with LLM providers (OpenAI, Anthropic, Mistral, open-source models via HuggingFace) . Design robust prompt templates, system instructions, and chain-of-thought strategies for reliable agent behavior . Implement RAG (Retrieval-Augmented Generation) pipelines connecting agents to structured and unstructured knowledge bases . Manage LLM output validation, structured output parsing, and fallback/retry logic Backend Services & API Layer . Build FastAPI-based microservices to expose agent capabilities as REST/streaming APIs . Develop event-driven agent triggers using message queues (Redis Streams, Kafka, RabbitMQ) . Design agent state management and session persistence using PostgreSQL / Redis . Implement task queuing, scheduling, and async execution for long-running agent workflows Tool & Data Integration . Build and register custom agent tools for file processing, web search, SQL querying, SFTP operations, and third-party API calls . Integrate agents with structured data sources (PostgreSQL, data lakes) and unstructured sources (PDFs, Excel, emails) . Connect agents to vector databases (Pinecone, Weaviate, pgvector, ChromaDB) for semantic retrieval . Enable agents to interact with external systems via MCP (Model Context Protocol) or custom tool registries Guardrails, Safety & Reliability . Implement guardrails and output validators to ensure agent responses meet business and compliance requirements . Build human-in-the-loop (HITL) checkpoints for critical decision nodes within agent workflows . Design agent sandboxing and execution boundaries to prevent unintended actions . Develop comprehensive error handling, retry strategies, and graceful degradation patterns for agent failures Observability & Evaluation . Instrument agents with end-to-end tracing (LangSmith, Arize, custom logging) across reasoning steps and tool calls . Build evaluation pipelines to measure agent accuracy, latency, tool-use correctness, and goal completion rates . Define and monitor KPIs for agent performance - task success rate, hallucination rate, escalation frequency . Maintain audit logs of agent decisions, tool invocations, and LLM interactions for debugging and compliance

Mandatory skill sets: Programming Python (advanced) - async programming, OOP, design patterns Agent Frameworks LangChain, LangGraph, AutoGen, CrewAI, or equivalent LLM APIs OpenAI API, Anthropic Claude, HuggingFace Transformers API Development FastAPI, REST, Streaming APIs (SSE / WebSockets) Prompt Engineering CoT, ReAct, structured outputs, function calling Data Handling Pandas, structured/unstructured data parsing Messaging & Events Redis, Kafka, or RabbitMQ for event-driven agent triggers Databases PostgreSQL for state/session management Observability Logging, tracing (LangSmith / OpenTelemetry), metrics

Preferred skill sets: Experience with fine-tuning or RLHF pipelines for domain-specific LLMs . Familiarity with Model Context Protocol (MCP) for standardized tool integration . Knowledge of open-source LLM deployment (Ollama, vLLM, LiteLLM) . Exposure to Kubernetes-based microservice deployment for agent services . Background in data engineering or ingestion platforms complementing agent data access . Understanding of AI safety principles, responsible AI, and model risk management . Experience with browser-use or computer-use agents for UI automation tasks

Years of experience required:

Education qualification: