Paras Parani

Graduate Research Assistant

Florida International University•  May 2024 - Present•  Miami, FL, USA

Fine-tuned pretrained Vision Transformers and LLMs, achieving a 15% accuracy improvement in seizure prediction. Designed lightweight transformer models for EEG data, improving accuracy by 5% over fine-tuned LLMs. Developing a multi-modal model for Alzheimer's detection using images and time-series data. Leveraged SLURM for distributed system management, reducing training time by 20%. Used Grad-CAM, SHAP, and attention map visualizations for explainability and detailed model analysis.

Senior Associate

Ignatiuz Software Pvt Ltd•  September 2021 - July 2023•  Indore, India

Led a team of 2 to successfully revamp the Scoutfoto project, delivering a more efficient and scalable platform. Developed the backend architecture and designed the database structure, optimizing data storage and retrieval. Designed and implemented scalable Django-based APIs, ensuring seamless integration and enhanced functionality. Deployed the platform on Azure, achieving an 80% improvement in operational efficiency and system reliability.