•Led a team of 5 to build a full-stack solution for converting 2D datasets into 3D models using Keras, PyTorch, and 3D Convolutional Neural Networks (CNNs), scaling the system to handle complex geometric shapes and improving 3D prediction accuracy by implementing mesh generation techniques.
•Utilized large language models (LLMs) and Natural Language Processing (NLP) techniques to convert text commands into 3D models, leveraging Transformer-based architectures for real-time text-to-3D model generation.
•Optimized machine learning workflows by integrating model pruning and hyperparameter tuning using scikit-learn and Optuna, reducing model generation time by 20%, and deployed the solution using Docker and Flask for containerized and scalable deployments.
•Automated data collection and preprocessing with advanced techniques such as data augmentation and synthetic data generation for training deep learning models, managing large-scale datasets with efficient pipeline automations.
•Developed a secure Flask-based frontend with JWT authentication and integrated Hugging Face BART and T5 models for automating text-tosummary processes, reducing manual efforts by 70%.
•Implemented model testing using PyTest for automated validation, and optimized the machine learning pipelines through techniques such as batch normalization and early stopping, reducing runtime by 20%.
•Managed end-to-end development with Git, Docker, and Jira, delivering the project on time and enhancing system reliability by 30%, while ensuring continuous integration and deployment (CI/CD) with Jenkins.