Deepak Kumar

About Deepak Kumar

Senior Data Scientist with 7+ years of experience specializing in Generative AI, Large Language Models, and Machine Learning. I build scalable, context-aware AI systems using LangChain, Langraph, Crew AI, AutoGen, Azure OpenAI, and Google Generative AI to drive automation, efficiency, and business impact. My expertise spans NLP, Prompt Engineering, RAG, predictive modeling, and multi-agent architectures.I focus on solving complex real-world problems by designing AI systems that integrate structured and unstructured data, deliver real-time insights, and automate decision workflows across domains like CPG, supply chain, and finance- Key Projects & Achievements1. GenAI-Powered Financial Insights Developed an LLM-based solution for investor call transcript analysis, extracting critical financial themes to enhance decision-making. Engineered RAG pipelines using Azure AI Search and vector databases for precise, semantic information retrieval.2. Multi-Agent Analytics Platform Built a Langraph-driven multi-agent system for sales, performance, and pricing analytics, integrating intent recognition, SQL query generation, and narrative insights. Deployed using Docker with Human-in-the-Loop (HIL) governance for transparent and scalable operations.3. Invoice-to-Cash (I2C) GenAI Automation – Collections, Cash Application & Deductions Developed a GenAI-powered, multi-agent Invoice-to-Cash application using Crew AI, streamlining three core modules: Collections Agent: Automated payment reminders, prioritized accounts based on risk, and generated collector insights using LLM-driven decision logic. Cash Application Agent: Automated remittance parsing, entity matching, and payment reconciliation by integrating LLM reasoning with enterprise data systems. Deductions Agent: Identified deduction types, extracted claim details, and generated resolution recommendations using GenAI-enhanced document understanding. This end-to-end system significantly reduced manual effort, improved cash flow visibility, and accelerated the I2C cycle through real-time automation and intelligent decision-making.**Technical AI & NLP: LLMs, Prompt Engineering, RAG, LangChain, Langraph, Crew AI, AutoGen, Azure OpenAI, Google Generative AI, Hugging Face*Data Science & ML: Regression, Classification, Text Mining, Predictive Modeling, Deep Learning*Tools & Platforms: Python, PySpark, SQL, Docker, Azure AI Search, Vector Databases*Domains: CPG, Supply Chain, Finance

Experience

1. Senior Data Scientist

   Fractal

   Apr 2024 — Present · Gurugram, IN

   Built a Langraph-based multi-agent system for sales, performance, and pricing analytics, leveraging collaborative agent workflows, intent recognition, and SQL query generation, delivering scalable, transparent insights and a 15% improvement in analytics turnaround time.2. Developed an LLM-powered solution using LangChain and Azure OpenAI to analyze investor call transcripts, extracting key financial insights and improving decision-making accuracy by identifying critical themes and trends.3. Designed real-time QA chatbots leveraging LangChain, Crew AI, and Retrieval-Augmented Generation (RAG), enabling context-aware responses from extensive document and transcript datasets, enhancing user query resolution by 30%.4. Engineered LLM-based document summarization tools with Google Generative AI and Hugging Face, improving information retrieval efficiency by 25% for lengthy reports and stakeholder presentations.

Education

• Dav public school pakur

  Intermediate science (CBSE)

  2011 — 2013

• GIET University Gunupur

  Bachelor of Technology, Electronics and Communications Engineering