Saurav Joshi Email & Phone Number

San Francisco Bay Area

Marina Del Rey, California, United States

Remote

Constructed a comprehensive novel dataset consisting of 7.7 million rows, leveraging rule-based heuristics and T5 open-source model, to aid in the translation of natural language questions into SPARQL queries over open knowledge graph DBpedia.Optimized the open-source Starcoder, CodeGen, and CodeLLama model via fine-tuning, we aim to secure a potential top spot on the KGQA leaderboard and currently working on a research paper

Los Angeles, California, United States

Constructed a data analysis pipeline at scale for efficient data transformation, modeling, and management.Developed a novel deep learning architecture along with statistical analysis that performs 3D MR vesselwall imaging using retrospectively acquired undersampled precontrast 4-min data and postcontrast 2-mindata to reduce the data acquisition time from 16-min to 6-min.

Mumbai, Maharashtra, India

Conducted intensive research on the efficiency of word embeddings in natural language processing on the IMDB movie review text dataset using Word2vec, GloVe, FastText, and BERT models.Developed a natural language processing pipeline that involved data gathering, data wrangling, data analysis, feature vectorization, data modeling, and model evaluation.Built a custom deep learning model using PyTorch framework for text classification purposes and wrote a paper - ‘A Study of Various Word Embeddings in Deep Learning’.

Mumbai, Maharashtra, India

Developed an end-to-end sales prediction application for a commodity firm, applied Information Gain, Chi-Square Test feature selection methods, and created an ARIMA model which improved the prediction accuracy by 12%.Constructed a Conversational AI Assistant for the Legal Industry, introducing new features namely multiple intent recognition, recommendation system that increased customer engagement and saved the company’s cost.Implemented a Few-Shot Learning Approach with Siamese Neural Network to overcome the problem of insufficient training data for the Conversation AI Assistant to improve the text classification accuracy and also designed data science and technology-related articles on Medium to help the community by sharing concepts and code.

Master'S Degree, Data Science, 3.85

Bachelor Of Technology - Btech, 9.43