Ram Korde

Contributed as a lead NLP engineer for text data preprocessing, cleaning and applying language models for various use-cases. Developed and trained transformer pipelines to create high-quality embeddings for client data, effectively building a comprehensive knowledge base for intelligent insights extraction. Applied BERT and zero-shot topic modeling techniques to create clusters for client's SDLC database. Leveraged LLMs and efficient embedding search to establish correlations in the knowledge base and retrieve accurate information to address user queries. Collected and curated data for model training and fine-tuning, ensuring the availability of high-quality and relevant datasets. Achieved a significant enhancement of 7-10% in correlation results through fine-tuning and hyperparameter optimization.

Developed and deployed highly effective machine learning models to forecast sales for AB InBev. Two of the models were identified as the top-performing solutions. Collaborated with a team of 4 Data Scientists to process and analyze sales data; and tune existing machine learning and statistical models. Achieved significant improvements in model accuracies, ranging from 5% to 10%. Employed Python and Excel for data visualization, effectively presenting sales history data and forecast results from the models. Shared insightful visualizations and analysis with clients to support decision-making processes.

Led an NLP initiative as part of the Robotic Process Automation team, focusing on automating the extraction of information from annual 10K forms of different companies. Developed a robust pipeline and a user-friendly GUI to process and extract information from over 20000 documents.