Our Research
We are dedicated to pioneering innovative solutions through cutting-edge research across diverse fields. By embracing interdisciplinary collaboration, we tackle real-world challenges and push the boundaries of possibility. Join us as we explore new horizons and create transformative impacts that enhance lives and drive societal progress.
MY PRECIOUS PII
Advancements in LLMs have prompted individuals and companies to fine-tune their own for specialized use cases. The million-dollar question is, when I fine-tune an LLM with my sensitive data (PII, PHI, FII, etc.), do I risk the tuned model leaking data or not?
​
​
PULSEFLEET
The Advanced Ambulance Routing System for Efficient Traffic Management is designed to optimize emergency vehicle navigation by integrating real-time traffic data, GPS, and predictive analytics. This system ensures the fastest and safest routes for ambulances, reducing response times and improving overall emergency response efficiency.
ArcGAN
This research suggests a hybrid architectural design model using generative adversarial networks (GANs) to merge old and new styles with high precision. By training on relevant data, the model simplifies the design process for non-experts while accurately reconstructing 3D structures from 2D images without requiring additional annotations.
AUTO-LVEF
This research introduces an interpretable AI model to calculate left ventricular ejection fraction from 2D TTE without human intervention. It addresses the explainability issue in healthcare AI and achieves 0.97 accuracy on the EchoNet-Dynamic dataset, outperforming current models.
​
​
GENOME
This research explores the intersection of genomics and cryptography, evaluating the potential, challenges, and methods of using human genome sequencing for cryptographic key generation. The human genome's complexity and uniqueness offer significant entropy, making it a promising resource for enhancing security systems.
DARKLIGHT
This research project, "Dark Light," is designed to assist visually impaired individuals using GPS and Arduino. It incorporates an ultrasonic sensor to detect obstacles within 2 meters, providing real-time audio feedback via headphones, while Google Assistant aids in precise navigation. The project aims to improve mobility and real-world interaction for visually impaired individuals.
I AM A BOT
The "I am a Bot" research focuses on developing an automated system that solves different CAPTCHA types using a multi-modal LLM. It identifies CAPTCHA formats (text, math, image-based) and provides solutions by leveraging Google's Vertex AI with Gemini Vision Pro for model inference. This research aims to enhance CAPTCHA-solving capabilities across various formats.