LLM​/GenAI Engineer

The role is responsible for moving LLM prototypes into resilient, production-grade systems, with a focus on advanced retrieval-augmented generation (RAG) architectures and multi-agent systems. The engineer will build mechanisms that ensure low-latency, deterministic, and highly accurate AI outputs for enterprise-grade applications.

This position collaborates closely with backend engineers, product managers, and data platform teams to integrate state-of-the‑art foundation models into existing production workflows. The ideal candidate cares deeply about observability, systematic evaluations, and the cost‑performance trade‑offs of modern GenAI systems.

• Design and optimize advanced RAG pipelines utilizing hybrid search, query rewriting, and reranking models to improve retrieval accuracy
• Build and maintain semantic search infrastructure across production vector databases, including Pinecone, Milvus, or Qdrant
• Implement systematic LLM evaluation suites and guardrails to monitor model outputs for hallucination, drift, and policy compliance
• Develop and deploy agentic workflows and tool‑calling architectures using Lang Chain, Lang Graph, or custom orchestration code
• Fine‑tune open‑source models (such as LLaMA or Mistral) using PEFT, LoRA, and QLoRA techniques for specialized domain tasks
• Optimize LLM inference pipelines for latency and cost using compilation frameworks such as vLLM, TensorRT‑LLM, or Triton Inference Server

• 3‑6 years of software engineering experience, with at least 1.5 years dedicated to building and deploying LLM‑based applications in production
• Expert‑level Python programming skills, including experience with asynchronous programming and building high‑throughput APIs
• Hands‑on experience with vector databases and structuring complex chunking and metadata strategies for unstructured data sources
• Familiarity with cloud‑native ML infrastructure (AWS, GCP) and containerization using Docker and Kubernetes
• Solid understanding of NLP and deep learning fundamentals, including transformer architectures and embedding spaces
• Bonus:
  Experience with model quantization, RLHF/DPO pipeline implementation, or contributing to open‑source LLM orchestration tools