About

Dipam Vasani is a highly skilled Data Science Engineer with a Master's degree in Data Science from Indiana University and a Bachelor's degree in Computer Engineering from Dwarkadas J Sanghvi College of Engineering. He is an experienced professional in Applied Machine Learning and Deep Learning with multiple projects and a Research Paper published in an International Conference. Dipam has hands-on experience with Data Analysis, Predictive modeling, Computer Vision, and Natural Language Processing (NLP) pertaining to his internships. He finished in the top 5% in the Intel Scene classification challenge and among the top 100 candidates in Yes Bank Hackathon. Currently, Dipam is working as a Data Science Engineer at Docusign, where he has previously worked as a Machine Learning Intern. He has used FasterRCNN with ResNet50 FPN as a backbone to detect various fields in documents, contracts, and forms. This model is 4x faster than the current model. He also trained Yolov3 on grayscale images for further speed boost. Dipam has trained an NLP model with custom embeddings and a Bi-directional LSTM on the text related to fields to classify the type of the field. This model improves the overall precision by 5%. The CV and NLP models are part of the pipeline for the auto-tagging feature. He has containerized Flask application using Docker and to run inference on test images using FasterRCNN and the NLP model. Dipam has also created a Gradio UI to easily test the NLP model on various test samples. He exported all models to ONNX for interoperability and faster inference. Dipam is also a technical writer for 7 different publications, including The Startup, Towards Data Science, Becoming Human, Heartbeat, Hackernoon, data-driven investor, and code heroku. He has a deep understanding of Data Science, Deep Learning, NLP, Computer Vision, ML, Research Scientist, test, Docker, Flask, and UI. With 1.94 years of relevant experience, Dipam is a highly motivated individual with exceptional problem-solving skills and attention to detail.