Mohit Gupta

Project: Finance Extraction Agent — Multi-Agent AI System. Built a multi-agent AI system to transform unstructured financial documents into US-GAAP–compliant, ledger-ready accounting records. Integrated LLM-based reasoning with deterministic GAAP validation rules to enforce accounting correctness, consistency, and auditability. Automated transaction classification and GAAP compliance checks, significantly reducing manual accounting effort.

Project: Conversational AI Chatbot for Policy Documents. Built a domain-specific chatbot using RAG architecture to answer queries related to organizational policy documents. Transformed PDF-based documents into structured Markdown using DocServe for efficient chunking and semantic search. Integrated ChromaDB as a vector database to store and retrieve document embeddings, enabling accurate and fast contextual search. Deployed the chatbot interface using OpenWebUI, offering a clean and interactive platform for end-users. Validated model performance using custom metrics—answer relevance, contextual accuracy, and precision—achieving 94% accuracy on benchmark datasets.

Project 1: Automated CVE Intelligence via Twitter Bot. Developed and deployed a serverless Twitter bot on Modal, scheduling tweets and AI-generated responses every 5 minutes. Project 2: Secure AI Agents & Custom GPT Development. Contributed to the development of enterprise-grade LLM solutions by collaborating on secure, precision-focused generative AI projects in the cybersecurity domain. Project 3: Latency Optimization in Regulatory Chat Search Pipeline. Pioneered system performance by reducing end-to-end latency from 12–15s to 6.3s (50%) through connection pooling, model preloading, and vector search optimization in Milvus. Project 4: Migration Project from Talend to Databricks using DLT. Improved query execution speed through the optimization of PySpark/Spark SQL workflows, facilitating a seamless migration from Talend to Databricks for large-scale data processing.