Veer Sanghavi

Experience

• 

  Tredence Inc.

  Dec 2022 - Apr 2023

  Data Science Intern

  Project Focus: Clinical text classification using Natural Language Processing (NLP).Objective: To categorize medical transcripts into corresponding medical conditions.Approach: 1. Employed Sequential Forward Selection (SFS) to streamline data dimensions and enhance classification performance.2. Preprocessed text data with tokenization, stemming, and stop word removal.3. Experimented with multiple machine learning algorithms including Logistic Regression, Support Vector Machines, and Categorical Boosting.Key Achievements:1. Achieved 99% accuracy in classifying two medical conditions using Logistic Regression.2. Identified the need for more data to improve accuracy for a broader range of medical conditions.Results: Demonstrated the effectiveness of SFS in optimizing the classification process and highlighted the importance of sufficient data for accurate predictions. Show less

• 

  Tech Mahindra

  Jan 2024 - Jul 2024

  AI Intern

  Detecting Synthetic (Fake) Human Face Images using advanced AI techniques:Framework Development: Developed a novel detection framework by leveraging StyleGAN2 inversion. This process involved extracting latent vectors from both AI-generated and real images, then regenerating the images to identify subtle differences.Dataset & Analysis: Worked with a robust dataset of 5,000 real images and 5,000 fake images. Applied several advanced image similarity metrics to compare the original and regenerated images, providing critical insights for detection accuracy.Model Training & Results: Trained Artificial Neural Networks (ANN) and XGBoost models on the similarity metrics, achieving near-perfect results with ~100% accuracy in distinguishing AI-generated images from real ones. Show less