Senior Manager, RAG/LLM Specialist

EXL; Gurugram District; On-site; INR 1,200,000 - 1,800,000; Full time

Job summary

EXL in India seeks a senior engineer to lead the design and optimization of advanced RAG pipelines and model fine-tuning for enterprise-scale LLM deployment.

You will own multi-stage pipelines, guide PEFT/LoRA tuning, develop evaluation methods, and mentor junior analysts in prompt engineering and coding quality.

Proficiency with Python, PyTorch/TensorFlow, LangChain, and LlamaIndex is essential to deliver low-latency, high-relevance results.

Qualifications

• Experience designing multi-stage RAG pipelines.
• Expertise in PEFT/LoRA fine-tuning for open-source models.
• Strong background in prompts optimization and embeddings for retrieval augmented generation.

Responsibilities

• Lead design and optimization of multi-stage RAG pipelines with low latency and high relevance.
• Drive fine-tuning initiatives (PEFT/LoRA) for open-source models to improve task performance.
• Develop automated evaluation frameworks to measure LLM accuracy, context precision, and recall.
• Architect vector database strategies and hybrid search (e.g., Pinecone, Milvus) for metadata filtering.
• Mentor junior analysts in prompt engineering, chunking strategies, and code quality.

Skills

Python

RAG pipelines

PEFT/LoRA

Prompt optimization

Embeddings

LLM deployment knowledge

Tools

LangChain

LlamaIndex

PyTorch

TensorFlow

Job description

Role Overview

Lead the design and optimization of advanced RAG pipelines and model fine tuning processes. Bridge the gap between prototype and enterprise-scale LLM deployment.

Key Responsibilities

• Pipeline Ownership: Design and manage complex, multi-stage RAG pipelines ensuring low latency and high relevance.
• Model Optimization: Lead fine-tuning initiatives (PEFT/LoRA) for open-source models to improve domain-specific task performance.
• Advanced Evaluation: Develop automated evaluation frameworks (e.g., RAGAS) to continually measure LLM accuracy, context precision, and recall.
• Vector Strategy: Architect metadata filtering and hybrid search strategies within vector databases (e.g., Pinecone, Milvus).
• Team Mentorship: Guide junior analysts in prompt engineering, chunking strategies, and code quality.

Required Skills

• Tech Stack: Python, PyTorch/TensorFlow, LangChain, LlamaIndex, advanced embedding models.
• GenAI Skills: Deep expertise in advanced RAG (HyDE, parent-document retrieval), prompt optimization, and parameter-efficient fine-tuning.