Job Title

Experience- 3 to 8 YearsLocation: PuneJob Description:We are looking for a hands on GenAI Engineer to design, build, and productionize AI solutions that leverage Retrieval Augmented Generation (RAG), Large Language Models (LLMs), LLMOps, and multi agent systems. You’ll own the end to end lifecycle—from data ingestion and orchestration to deployment, evaluation, guardrails, and monitoring—while collaborating with product, platform, and domain teams to ship reliable, cost efficient AI features at scale.What You’ll Do (Key Responsibilities)Solution Architecture & DeliveryDesign RAG pipelines (chunking, indexing, embeddings, retrieval, re ranking, synthesis) and select optimal vector DBs & re rankers for each use case.Build multi agent workflows (planner/executor, tool using agents, collaborative agents) with robust state management and failure recovery.Implement LLMOps: automated evaluations, data & model lineage, observability, cost controls, rollback strategies.Data & Retrieval EngineeringBuild ingestion pipelines for unstructured/semi structured content (PDFs, Office docs, HTML, emails, logs) with robust parsing, PII redaction, deduplication, and metadata enrichment.Optimize embeddings (model selection, dimensionality, multilingual handling) and retrieval quality (query transformation, hybrid search, learning to rank).Prompt Engineering & SafetyDevelop system prompts, tool calling schemas, and guardrails (content, privacy, compliance) with iterative prompt optimization and A/B testing.Implement safety, policy, and governance: jailbreak resistance, hallucination mitigation, citation enforcement, rate limit handling.Orchestration & DeploymentProductionize pipelines using workflow engines (e.g., LangGraph, Autogen, CrewAI, Airflow/Prefect) with containerization (Docker) and Kubernetes.Deploy & scale inference (vLLM, Triton, Ray, Helm) across cloud/on prem; manage secrets, keys, and per tenant configs.Evaluation, Monitoring & Cost OptimizationDefine and track metrics: answer correctness, faithfulness, groundedness, latency, throughput, token cost, retrieval observability (Prometheus/Grafana), tracing (OpenTelemetry), and E2E analytics (W&B/MLflow) with automated regression tests.Tech Stack (You don’t need all; we value depth in relevant areas)LLM & RAG: OpenAI/Azure OpenAI, Anthropic, Google, Cohere; open source models via Hugging Face (Llama, Mistral, Qwen).RAG Components: FAISS, Milvus, PineconeFrameworks: LangChain, LlamaIndex, LangGraph, Autogen, CrewAI; tool calling & function schemas.Orchestration & Pipelines: Airflow, Prefect, Dagster; Ray for distributed workloads.Serving: vLLM, Triton Inference Server, FastAPI, gRPC; Helm/K8s, Istio/Linkerd.Ops & Observability: MLflow, Weights & Biases, OpenTelemetry, Prometheus/Grafana; Feature/Model Registry.Data & Parsing: Unstructured, Apache Tika, Textract, Tesseract; Pandas/Spark.Caching & Messaging: Redis, Kafka.Qualifications (Must Have)Solid hands on experience building with LLMs and RAG, including retrieval tuning and prompt/system design.Proven ability to productionize GenAI workloads (Kubernetes, CI/CD, containerization, secrets, autoscaling, rollout/rollback).Experience with LLMOps: evaluation frameworks (e.g., RAGAS/DeepEval/HELM style metrics), tracing, monitoring, and cost management.Strong prompt engineering expertise: tool calling, schema design, structured outputs, guardrails, and prompt A/B testing.Exposure to multi agent systems (planner/executor, tool orchestration, memory/state) and their failure modes.Proficiency in Python (typing, testing, packaging) and building APIs/services (FastAPI) with clean architecture patterns.Understanding of data privacy, security, governance in AI systems (PII handling, policy enforcement, model risk).Nice to HaveExperience with vector DB benchmarking and embedding model selectionDocument intelligence (layout aware parsing, table extraction, OCR quality improvement).Domain exposure to SDLC automation (code review assist, test generation, requirement traceability) or manufacturing/agri OEM knowledge.Contributions to open source GenAI projects; publications or talks a plus.