Machine Learning Engineer III (gen AI & Multi-agentic Systems)

Expedia Expedia · Hospitality · Bangalore, India

Machine Learning Engineer III at Expedia Group focused on building end-to-end GenAI products, including agent orchestration, inference platforms, and user-facing interfaces. The role involves LLM training, fine-tuning, RAG pipelines, agentic workflows, and implementing evaluation and safety frameworks, with a focus on scaling these systems for consumer travel applications.

What you'd actually do

  1. Architect, implement, and scale LLM training, fine-tuning, adaptation (LoRA/QLoRA/adapters), and distillation pipelines, including RLHF/DPO for production GenAI systems and chatbots.
  2. Build and optimize RAG pipelines with vector stores and retrieval memory layers, supporting multimodal data such as vision, audio, text, and music.
  3. Design and develop agentic workflows using frameworks like LangChain, AutoGen, LangGraph, and OpenAI Agents SDK, including typed tool contracts, schema validation, and retry logic.
  4. Implement evaluation, safety, and reliability frameworks using LangSmith, DeepEval, and OpenClaw, including automated metrics, hallucination mitigation, and LLM-as-a-Judge evaluation.
  5. Build and deploy production inference platforms optimized for latency and cost using vLLM, TensorRT-LLM, and DeepSpeed.

Skills

Required

  • Python
  • PyTorch
  • Docker
  • AWS
  • Azure
  • LangChain
  • AutoGen
  • LangGraph
  • OpenAI Agents SDK
  • MLflow

Nice to have

  • vLLM
  • TensorRT-LLM
  • DeepSpeed
  • LangSmith
  • DeepEval
  • OpenClaw

What the JD emphasized

  • production ML/GenAI systems
  • LLM fine-tuning, adaptation, distillation, RLHF/DPO
  • RAG and multimodal systems
  • agent development and orchestration
  • delivering end-to-end GenAI products
  • UX/UI design principles for AI systems and chatbots
  • context engineering

Other signals

  • build complete end-to-end AI products
  • agent orchestration
  • inference platforms
  • user-facing interfaces
  • LLM training, fine-tuning, adaptation
  • RAG pipelines
  • agentic workflows
  • memory strategies
  • evaluation, safety, and reliability frameworks
  • production inference platforms
  • cloud-ready systems
  • CI/CD, experiment tracking, and model versioning
  • model drift, agent performance, and tooling reliability
  • translate requirements into measurable, user-facing GenAI products
  • intuitive user interfaces and experiences for agentic systems