My projects makes use of vast variety of latest technology tools. My best experience is to create Data Science projects and deploy them to web applications using online hosting infrastructure.
Developed and refined Global Dispute and Escalation Prediction AI model with 75%+ F1 score, resulting in improved efficiency for Finance Collectors in resolving payment delays, to deploy on AR Cloud and Engage AR.Leveraged Watsonx.AI Large language models to extract Contract Accounts Receivable terms in hackathon estimated 10M+ impact.
The PDF AI Bot is a Streamlit app that integrates Large Language Models, Langchain, and Weaviate vectorstore to interactively answer questions from uploaded PDFs using natural language processing and vector search technology.
Columbia Data Science Dev Hackathon 2023: Led in a team of 4, analyzed contributions to the disparities in serious adverse outcomes of the various drug treatments approved by FDA ( Federal Drug and Food Administration). Specifically, we were curious about certain drugs, the impacts of sex, patient weight, patient onset age, and the drug's route of administration, Learnt about API limits, faced issues in data collection, and scalability, and understood the up-sides of building something that can impact lives.
Machine Learning solution to classify Software Licenses, given Software Support Descriptions automatically as part of NPI (New Product Intro) process saving 95% time. Experimented with 220+ finetuned Model experiments using LSTM & Neural network based optimized architectures.