Sparsh Srivastava

AUM Exchange: Scraping property details using Selenium and Beautifulsoup. Generating property and locality descriptions using OpenAI's GPT models with Langchain. Extracting named entities (Names, Organizations, Addresses, Monetary values) from loan agreements using AWS Comprehend. Created and deployed a locally deployed chatbot using GPT4ALL with langchain and RAG capabilities on an AWS instance. All tasks performed using Object Oriented Programming techniques. Details Extraction and Description: Used GPT with langchain to extract details about patterns from portals like Etsy and PDF magazines (using PDFminer and pytesseract). Generated bespoke descriptions using LLM chains by Langchain. Created and deployed a Docker container for this capability on an AWS instance. Property Price Prediction: Used data including ratings, locality information, amenities, and price/sq ft (target variable).

Verbatim Classification (NLP): Automated the classification of verbatim call accounts into various sentiment levels (e.g., General Comment, Customer Service Challenges). NPS (Net Promoter Score) Analysis: Performed analysis on customer experience metrics. D-Commerce Foundational Analytics: Processed and harmonized Sales, Ads, and Customers data from three different retailers for dashboard generation in a BI tool.

Assessment Independent Counselling: Developed a system that suggests career clusters, pathways, and job roles based on user responses, classifying responses using John Holland's Theory of Career Choice (RIASEC) and employing NLP models like XLnet and BERT. Recommender System: Implemented both collaborative filtering (for enrolled courses) and content-based recommender systems (using embeddings generated from course descriptions, specializations, and skills) to recommend courses, further sorting recommendations based on weightage and mapping required educational qualifications.