Shubham Pandey

Developed a medical document processing pipeline using Visual Language Models, extracting tabular data and key-value pairs with 40% improved efficiency through advanced layout parsing. Built a proof-of-concept web app for KPMG with GPT-4o-mini and LangChain, enhancing internal query handling by rephrasing inputs based on historical context, reducing SQL query errors by 25%.

Designed an automated driving test system using IP cameras and real-time path tracing, cutting manual evaluation effort by 75% and saving an average of 700k (In INR) annually in operational costs. Engineered an interview platform with GPT-3.5 and RAG, generating job-specific questions from input parameters, slashing candidate shortlisting time from 1 week to 2 days.

Developed and implemented multi-modal (NLP+Vision) models for document classification, parsing, and information extraction from medical and KYC documents, achieving high accuracy in NER, POS tagging, and summarization tasks. Deployed models as scalable Flask APIs on AWS EC2 with Docker, ensuring efficient model serving.

Collected market and accounts data from various sources, including IBES, WRDS for Compustat, and CRSP. Developed an Earnings Forecast model to predict earnings per share for NYSE and NASDAQ listed firms, leveraging accounting and trading data.

Grade: 7.91

Studied Computer Science and Engineering, gaining foundational knowledge and practical skills in the field. Achieved a GPA of 7.91, demonstrating strong academic performance.

Grade: 8.14

Completed Intermediate studies with a focus on core subjects. Achieved a GPA of 8.14, reflecting dedication and academic achievement.