Our client, a venture-backed AI Startup, is hiring a talented ML/AI Research Engineer to join their team in San Francisco. The successful candidate will lead the design, training, evaluation and optimization of agent-native AI systems, working at the cutting edge of LLMs, vector search, graph reasoning and reinforcement learning to build the intelligence layer on top of their enterprise data fabric.
Responsibilities
• Fine-tune and evaluate open-source LLMs (e.g. LLaMA 3, Mistral, Falcon, Mixtral) for enterprise-grade applications.
• Build and optimize RAG pipelines using tools such as LangChain, LangGraph, LlamaIndex or Dust.
• Develop and iterate on agent architectures (ReAct, AutoGPT, BabyAGI, OpenAgents) using real-world enterprise workflows.
• Design embedding-based memory systems with efficient, high-performance retrieval strategies.
• Implement reinforcement learning pipelines (RLHF, DPO, PPO) to improve agent behavior and decision-making.
• Create scalable evaluation frameworks, including synthetic evaluations, trace capture and explainability tooling.
• Own model observability, drift detection and alignment strategies across production systems.
• Optimize inference latency and GPU utilization across cloud and on-premise infrastructure.
Skillset
• Strong experience fine-tuning open-source LLMs using frameworks such as HuggingFace, DeepSpeed, vLLM, FSDP and LoRA/QLoRA.
• Hands-on experience with modern alignment techniques, including SFT, RLHF and DPO pipelines.
• Proven ability to build high-quality training datasets and robust evaluation frameworks for LLM systems.
• Deep understanding of scaling and optimization trade-offs, including batching, context windows, precision and quantization.
• Experience building and deploying production-grade RAG systems.
• Familiarity with orchestration and retrieval tools such as LangChain, LangGraph and LlamaIndex, and vector databases (Weaviate, Qdrant, FAISS).
• Experience working across structured (SQL, graph) and unstructured data sources.
• Experience designing agent-based systems with memory, tool use, and multi-step reasoning.
• Strong understanding of agent workflows (e.g. Plan-Act-Reflect), including self-correction and multi-agent systems.
• Expertise in inference and retrieval optimization, including chunking strategies, reranking, and low-latency deployment (e.g. vLLM, TGI).
Benefits
• Salary: $180k - $240k
• Equity.