Priyam M
- Research Program Mentor
PhD candidate at University of Illinois at Urbana Champaign (UIUC)
Expertise
Artificial Intelligence, Computer Vision, Natural Language Processing, Neuroscience
Bio
Hey everyone! My name is Priyam and I am currently a PhD student at the University of Illinois Urbana-Champaign in Electrical Engineering. More specifically I am studying applications of Artificial Intelligence in Speech Processing! I am essentially a Neuroscientist and Statistician that has fallen in love with AI and I aim to build Machine Learning Architectures that better mimic the human learning process. My knowledge has become a toolbox and I am able to apply these skills to bring state-of-the-art methods to problems nobody has thought of before. Outside of my research I am an avid wildlife photographer. I spend most of my time hiding in bushes waiting for a critter to come by so I can take incredible photos of them. Photography was my main motivation initially to dig into AI and I hope that I can continue to build tools that allow me to grow my Art!Project ideas
Multimodal Artificial Intelligence
Most Deep Learning models are task oriented and can specialize in one thing. This could be Image Classification, Text Generation, etc... The problem with this is the human mind doesn't work as a set of discrete processes but rather a large cohesive system. Multimodal AI is an advanced area where our goal is to see how we can merge together Images, Text, and Sound all together in a single model to enable even more human-like capabilities in the future.
Automatic Speech Recognition
Voice to Text is an important tool needed especially for users with disabilities. Often to interact with technology, people have no option but to give commands or transcribe their speech, meaning it is crucial to build a robust speech recognition system. There are extremely advanced methods for this utilizing Transformers, but first learning to build such a system and testing it on different use cases may offer some intuition for future work.
Geospatial AI
We have had consistent and accurate data collected about the Earth's surface for more than 3 decades now, but only in this new era of Big-Data are we able to leverage it. Understanding the qualities of the earths surface, detecting specific environmental features, and classifying the land use of different parts of the earths surface, is crucial to do in realtime. Computer vision has offered many techniques that can be explored to perform some of these tasks.