I am a PhD candidate in Information Science at Cornell Tech in New York City. My PhD research is on Large Scale Data Mining for Healthcare. Additionally I have a strong interest in data-mining/machine-learning applied to cyber-security & video analytics. During summer 2016, I interned at Dropbox with Abuse Prevention team. Before PhD, I completed Master of Engineering in Computer Science from Cornell.
I take a unique approach to research, which involves building systems that serve a practical purpose and can be quickly deployed to produce significant impact. I am currently working on two systems Computational Healthcare & Deep Video Analytics.
Today medical researchers can acquire de-identified data on millions of patients. Yet currently there is no system that allows them to quickly explore & use this data while ensuring compliance with privacy requirements. Computational Healthcare platform indexes and aggregates data from millions of patient visits, enabling physicians and researchers to conduct high-impact medical research.
Computational Healthcare is developed in collaboration with Radiology, Anesthesiology & Neurology departments at Weill Cornell Medical College. It is open source and available for use by researchers.
I am currently working on Deep Video Analytics an open source video analytics platform. The platform uses deep learning based indexing, detection and recognition models for visual search. Its highly customizable and can be used for building complex applications such as Dash Cam Analytics.
Dash Cam Analytics: Videos recorded by dash cams of personal vehicles, taxis, etc. not only provide information about the cars and drivers but also about roads and surroundings. Information derived from these videos has potential to provide immense value. Dash Cam Analytics is a platform for analyzing these videos.
I have developed a simple visual indexing and search system, using features derived from Google's inception model trained on the imagenet data. Images are efficiently (2$ per 1,000,000 images) indexed using AWS spot GPU instances. An Approximate Nearest Neighbor query server is used for efficient querying.
To experiment with developing interactive computer vision applications in JS, I built EraseImage.com. Its an online tool for segmenting images, that lets users perform image segmentation and background removal completely in client-side JS. It is implemented using Angular, FabricJS and superpixel algorithms.
Copyright Akshay Bhat, 2017. All rights reserved.