Harshit Kansal

I’m a GenAI-focused AI Engineer specializing in building scalable LLM, RAG, and agent-based AI systems for enterprise automation using modern AI frameworks and Azure AI. I like turning complex AI workflows into reliable, reusable components that teams can actually deploy and govern.

At NTT Data (Dec 2024–Present), I enhanced NeuroStack, a modular agentic AI platform on Azure, by building LLM and AI agent workflows for enterprise decision automation. I developed RAG pipelines in Python using chunking, embeddings, vector search, and prompt orchestration, then designed knowledge-base ingestion flows to process documents, create embeddings, and store searchable context in vector databases.

I also engineered real workflow automation: digital claim intake with Azure Document Intelligence and RAG, reducing manual document review effort by 30%, and customer screening for EQ Bank, improving compliance workflow efficiency by 25% with LLM-assisted decision support. I built reusable LangChain and LangGraph components for retrieval, reasoning, workflow routing, and agent execution—tightening prompt templates and retrieval logic to increase answer relevance and reduce hallucination risk.

I’m committed to safe, observable AI in production, so I implemented AI Governance and AIOps workflows with ServiceNow monitoring to improve model compliance and operational visibility. Earlier, I built an “AI Fraud Detection System” using Python, LLMs, and RAG, which reduced manual fraud review time by 35%, and I’ve been recognized with a Star Award at NTT Data for strong ownership and contributions to AI workflow delivery.