Ishaan Roy

Background

About Ishaan Roy

Welcome! I'm a seasoned machine learning engineer passionate about pushing AI boundaries through deep learning in computer vision. With over 3 years' experience, my expertise spans Python, Signal and Image Processing, and Deep Learning. Let's innovate together! Key Highlights: I’m currently pursuing a master’s degree in electrical and computer engineering (Signal and Image Processing): •Research in LLMs: Spearheaded the development of an automated framework to detect hallucinations in large language models (LLMs). This does not make use of any external database or human verification. • Generative modelling in vision: Developed, trained, and evaluated a model based on the stable diffusion to generate continuous 3D consistent views of indoor scenes using textual prompts. • Successfully completed courses in advanced probability theory, deep learning in computer vision, advanced neural networks, and machine learning systems engineering. Before rejoining academia, I worked in the medical technology space (Sleepiz AG) for two years as a signal processing and machine learning engineer: •‍ Algorithm Development: Designed, developed, and executed several deep learning projects to predict health metrics with medical grade accuracy (> 95% accuracy). • Dataset Management: Curated large complex multidimensional datasets acquired from more than 1500 live devices, spread across time zones and deployed in diverse environments. •🧠 Transfer Learning: Developed the first generation of transfer-learning based algorithms to expand the Sleepiz ecosystem to other novel devices and without experiencing any deterioration in algorithm performance. I also briefly interned at Samsung Research in the core networks and communications team, where I explored in leveraging AI in 5G systems: •🤖 Aided in the development and implementation of a deep reinforcement learning algorithm for dynamic radio resource allocation, contributing to the beyond 5G initiative. • Led the preparation and publication of a large-scale dataset on Samsung's internal data warehouse, comprising over data points. • Led the development of GAN-based algorithms for predicting radio maps from raw skymaps, optimizing accuracy and efficiency. Skills: Python, Deep Learning, MLOps, Signal Processing, AWS, Azure Machine Learning, Docker, CI/CD, Git, MySQL Let's connect and collaborate to drive innovation and create meaningful impact together. Looking forward to hearing from you!