Headline

AI @KPMG | Search & Retrieval · RAG · LLM Fine-tuning · Multi-Agent Systems | Research -​> Product | IEEE Published @ CAI 2026 | MTech IIIT Allahabad | Applied AI for Production

About

I'm an AI Engineer at KPMG building production-scale generative AI and RAG systems that solve complex enterprise problems. My Expertise: LLM fine-tuning (LoRA, QLoRA), retrieval-augmented generation with semantic search, multi-agent system orchestration, finding business gap and delevering ML solutions and transforming cutting-edge research into deployment-ready software. My Journey: I completed my M.Tech in Information Technology (Machine Learning) at IIIT Allahabad, before that worked at Cognizant on data science and NLP problem-solving, and now at KPMG I develop end-to-end AI solutions, while actively contributing to open-source projects in the AI/ML space.​ What I'm Building: I believe production AI isn't just about accuracy—it's about interpretability, scalability, and responsible deployment. I share insights on LLM optimization, RAG best practices, and lessons learned scaling AI systems from research to production. Open to: Collaborating on open-source AI projects, technical discussions about AI/ML systems, and connecting with engineers solving similar problems at scale.

Location

Bengaluru

AI stack

claude code

claude

hugging face

langchain

langgraph

langsmith

Work experience

AI Engineer at KPMG, Jul 2025 - Present · 1 yr 3 mos, Bengaluru, Karnataka, India

AI Intern at KPMG, May 2025 - Jun 2025 · 2 mos, Bengaluru, Karnataka, India

Project Research Associate at Indian Institute Of Information Technology Allahabad, Nov 2023 - Jul 2025 · 1 yr 9 mos, Vision Transformers have transformed how we approach computer vision, yet a subtle limitation persists, attention mechanisms across layers operate independently, restricting how information flows through the network. In HAViT, we tackle this with a cross-layer attention propagation method that preserves and integrates historical attention matrices across encoder layers, enabling progressive refinement of attention patterns throughout the transformer hierarchy. Preprint: https://lnkd.in/gUBcfFuy Code: https://lnkd.in/gYqGMWQ8

Senior System Engineer at Cognizant, May 2022 - Sep 2023 · 1 yr 5 mos, Kolkata, Built production ML pipeline processing 100K+ insurance claims using PySpark -> reporting time reduced by 20 hours/week Developed ensemble prediction system (XGBoost/LightGBM) for 5K+ daily claims with A/B testing -> manual review reduced by 40% Engineered BERT-based NLP pipeline for NER on 50K+ unstructured documents -> extraction accuracy +12%

Education

Master of Technology - MTech, Information Technology(Machine Learning), Indian Institute Of Information Technology allahabad

Bachelor of Engineering - BE, Electronics and Communication Engineering, Academy of Technology