Job Description

• Design and implement LLM-based agents using frameworks like LangGraph, LangChain, and custom agent stacks with support for memory, tools, and multi-step reasoning.
• Build and orchestrate complex agent workflows, including tool-use chains, self-correcting agents, and collaborative agent systems.
• Develop and optimize Retrieval-Augmented Generation (RAG) pipelines, including Graph RAG and hybrid retrieval techniques.
• Conduct end-to-end evaluation of LLM-based systems, including agent performance, retrieval relevance, grounded-ness, generation quality, faithfulness, and hallucination rates.
• Use tools such as DeepEval, RAGAS, LangSmith, and custom metrics to assess and optimize LLM outputs and agent decision-making.
• Apply Prompt Engineering strategies, including task-specific prompting, few-shot patterns, CoT reasoning, and tool-calling formats.
• Use PEFT techniques like LoRA, QLoRA, and RAFT to efficiently adapt models to specific domains.
• Collaborate with cross-functional teams to integrate LLM capabilities into scalable products and systems.

Job Requirements

• Over 5 years of experience in NLP systems and applied machine learning, with a minimum of 3 years focused on developing AI agents and retrieval-augmented generation (RAG).
• Extensive experience in designing agent-based systems utilizing LangGraph and other orchestration tools (such as LangChain and custom agent frameworks).
• Knowledge of evaluation methodologies for large language models and RAG, encompassing groundedness, faithfulness, and retrieval efficiency.
• Strong backend development skills in Python.
• Expertise in Prompt Engineering, including system prompting, few-shot and zero-shot configurations, chain-of-thought prompting, and tool utilization.
• Practical experience with parameter-efficient tuning methods (like LoRA, QLoRA, RAFT) for customizing base models to particular tasks.
• Proficient in Python as well as deep learning frameworks (including PyTorch, TensorFlow, and HuggingFace Transformers).
• Awareness of retrieval systems and vector search technologies (such as FAISS, Milvus, and Qdrant).

Skills