Looking for a job

Saurabh Srivastava

@saurabhsrivastava2

Message

I build production agentic AI and advance video scene graph reasoning.

India

Message

What I'm looking for

I'm looking to bridge rigorous AI research with real-world impact by building production-grade agentic AI, LLM-powered systems, and computer vision solutions.

At Naukri.com, I build production-grade agentic AI for intelligent job search, including the Job Seeker Agent. I design intent classification and entity extraction pipelines, fine-tune open-source models with LoRA and SFT, and build LLM evaluation systems for production queries.

My M.S.(R) research at IIT Delhi advances scene graph anticipation, spatio-temporal reasoning, and video understanding. My work has appeared at ECCV 2024, CVPR 2025, and CVPR 2026, including the introduction of Scene Graph Anticipation and an ECCV oral presentation.

Experience

Work history, roles, and key accomplishments

Current

AI Scientist

Current

Info Edge (Naukri.com)

Jul 2025 - Present (1 year 2 months)

Building agentic AI systems for job search and career assistance, including the Job Seeker Agent. Designed intent classification and entity extraction pipelines, fine-tuned models with LoRA and SFT, and built an LLM evaluation framework.

Agentic AI LLM Intent Classification Entity Extraction LoRAs SFT LLM Evaluation LLM As A Judge Prompt Engineering Hugging Face VLLM MCP Python

Associate Consultant

PricewaterhouseCoopers (PwC)

Jul 2019 - Apr 2021 (1 year 9 months)

Built and deployed ETL pipelines for a Tier-1 banking client using SQL and SAP BODS, consolidating Call Detail Records into a centralized Data Warehouse to support analytics and strategic decision-making.

ETL SQL BODS Data Warehouse Data Analysis

Education

Degrees, certifications, and relevant coursework

Indian Institute of Technology Delhi

M.S.(R), Computer Science and Engineering

2022 -

Grade: 8.26

Pursuing M.S.(R) in Computer Science and Engineering with thesis submission expected in 2026. Research focuses on scene graph anticipation and robustness for activity recognition.