Piyush Agrawal
Computational Neuroscience | Natural Language Understanding (the real kind) | Senior AI Scientist @ Rightindem
London, England, United Kingdom
645 followers
500+ connections
About
Primarily, my research interests lie in Understanding i.e. how neural perception works with applications in NLU, Computational, and Cognitive Neuroscience. My long-term goal is to research how we understand the world around us and explore if artificial perception models can be created for the same.
My aim is to create language models that, exhibit a level of generalization similar to that of humans and align more closely with human intuition with a moderate amount of data. Research on language comprehension is a great prospect for discovering how our internal learning and understanding mechanisms work as the human brain can efficiently decode any language with very low computational cost and resources. To achieve this, I work on developing computational cognitive models that mimic human language comprehension. A true artificial perception model for NLU may help to bridge the gap between human and machine communication. We have progressed in “processing” but have a long way to go in the “understanding” aspect of natural language in artificial systems.
It would be absolutely incredible to know if there's a universal perception algorithm that our brains use!
Activity
-
Watch our video to see what a small part of one of our AI modules can do to assist you and your claimants with unstructured data! Everyone's…
Watch our video to see what a small part of one of our AI modules can do to assist you and your claimants with unstructured data! Everyone's…
Liked by Piyush Agrawal
-
I'm delighted to announce that Christopher Brown is now the new Chief Technology Officer at RightIndem. His appointment is a pivotal moment for our…
I'm delighted to announce that Christopher Brown is now the new Chief Technology Officer at RightIndem. His appointment is a pivotal moment for our…
Liked by Piyush Agrawal
-
NEW: Generative AI is getting kicked off its pedestal — it will be painful but it’s not a bad thing It’s been two years since the phrase “generative…
NEW: Generative AI is getting kicked off its pedestal — it will be painful but it’s not a bad thing It’s been two years since the phrase “generative…
Liked by Piyush Agrawal
Experience
Education
-
University of Sussex
Master of Science - MS Artificial intelligence and Adaptive Systems First Class with Distinction
-
Activities and Societies: Data-intensive Science Center University of Sussex (DISCUS)
-
Savitribai Phule Pune University
Bachelor of Engineering - BE Computer Engineering 8.77 (First Class with Distinction)
-
Activities and Societies: CSI (Computer Society of India), Techtonic (2019, 2020)
Licenses & Certifications
Publications
-
Can humans help BERT gain "confidence"?
The advancements in artificial intelligence over the last decade have opened a multitude of avenues for interdisciplinary research. Since the idea of artificial intelligence was inspired by the working of neurons in the brain, it seems pretty practical to combine the two fields and take the help of cognitive data to train AI models. Not only it will help to get a deeper understanding of the technology, but of the brain as well. In this thesis, I conduct novel experiments to integrate cognitive…
The advancements in artificial intelligence over the last decade have opened a multitude of avenues for interdisciplinary research. Since the idea of artificial intelligence was inspired by the working of neurons in the brain, it seems pretty practical to combine the two fields and take the help of cognitive data to train AI models. Not only it will help to get a deeper understanding of the technology, but of the brain as well. In this thesis, I conduct novel experiments to integrate cognitive features from the Zurich Cognitive Corpus (ZuCo) (Hollenstein et al., 2018) with a transformer-based encoder model called BERT. I show how EEG and eye-tracking features from ZuCo can help to increase the performance of the NLP model. I confirm the performance increase with the help of a robustness-checking pipeline and derive a word-EEG lexicon to use in benchmarking on an external dataset that does not have any cognitive features associated with it. Further, I analyze the internal working mechanism of BERT and explore a potential method for model explainability by correlating it with a popular model-agnostic explainability framework called LIME (Ribeiro et al., 2016). Finally, I discuss the possible directions to take this research forward.
-
Manipulating muscle activity data from Electromyography for various applications using artificial intelligence
2021 Springer International Conference on Computational Intelligence | Algorithms for Intelligent Systems
Best paper award in track IOT, Image/Signal Processing (ICCI 2021)
Electromyography is used to capture electrical signals produced by skeletal muscle fibers. It is widely used in the medical field for evaluating problems in the nerves and muscles. It has been proven to provide accurate results that can even differentiate movements on a millimeter scale. New applications for EMG are being researched in numerous sectors. Uniqueness in any form of data is a boon for AI applications, and…Best paper award in track IOT, Image/Signal Processing (ICCI 2021)
Electromyography is used to capture electrical signals produced by skeletal muscle fibers. It is widely used in the medical field for evaluating problems in the nerves and muscles. It has been proven to provide accurate results that can even differentiate movements on a millimeter scale. New applications for EMG are being researched in numerous sectors. Uniqueness in any form of data is a boon for AI applications, and muscle activity generates a unique pattern of electrical signals for different movements like hand pinch or flexing. Artificial intelligence can be leveraged for the data generated by the human body for classification, prediction, and or analysis. Pairing the two together may lead to applications in diverse fields. In this paper, we experimented with applications like detecting motor neuron disease and hand gesture recognition using artificial intelligence with EMG data. We show improvement over the results of previous studies on gesture recognition and ALS detection using one-dimensional time series classification and CNNs. -
Survey on peer-to-peer ridesharing for "Pool" a ride-sharing app
International Journal of Engineering Applied Sciences and Technology
A review article for my app Pool.
Many college students travel in public transports or walk a long distance to reach college. This is problematic because public transports can be slow and not available everywhere as they have a specific time of arrival in their stops and they have to halt at multiple places in the city which can make it quite time consuming for passengers to reach their destinations. The goal of our project is to reduce this problem by providing a ride sharing application…A review article for my app Pool.
Many college students travel in public transports or walk a long distance to reach college. This is problematic because public transports can be slow and not available everywhere as they have a specific time of arrival in their stops and they have to halt at multiple places in the city which can make it quite time consuming for passengers to reach their destinations. The goal of our project is to reduce this problem by providing a ride sharing application for institutes. This will be mutually beneficial for the students providing a ride and the students wanting to reach their destination quickly and
cheaply as those who bring their own vehicles anyhow have to go to their homes without anyone sharing the ride with them. This will help them to earn money to at least cover their transportation or fuel cost and in-turn help provide a cheap ride to the ones in need. In this paper, we survey the work that deals with various paradigms of ride sharing and coincides with our idea for the application.Other authorsSee publication -
Extending drone’s capabilities for autonomous flight approach combined with indoor pod delivery mechanism
International Journal of Future Generation Communication and Networking
This paper focuses on the tools and algorithms available today and how they can be extended or modified to construct an indoor delivery system by combining delivery drones and a network of indoor tubes.
Projects
-
Teaching assistant for Artificial Intelligence
-
Developed a chatbot using an Encoder-Decoder Transformer model from scratch to teach about artificial intelligence with a curated dataset.
-
Motor Neuron Disease detection using artificial intelligence
-
-
Laser Tag
-
Led a team for the development of a fully-featured laser tag game with programmed vests and guns with live counters and multiple game modes to be hosted in the annual tech event at my college.
-
Electronic Drumming Kit
-
Made an electronic drumming machine using Piezoelectric Sensors and MIDI with a micro-controller.
-
(B2C) Niche Mobile-Commerce App
-
Made a mobile commerce app for dairy shopping and accounting for a dairy farm.
-
Trac
-
A voice-controlled material design music player.
-
HoloBox
-
An interactive gesture-controlled device capable of displaying real-time 3D hologram projections.
-
Home. — Home Simplification Device.
-
A device which uses an app or voice commands (NatLI) to operate consumer appliances. It does not require any smart devices to work. Just the average appliances. Works over the internet so you can control your home wherever you are. With support for the interoperability of different IoT devices.
-
NatLI (Natural Language Interface) — Assistant for PC
-
A voice-enabled NLP assistant with deep integration and automation of Windows OS.
-
The Board.
-
An inexpensive large interactive multi-touch screen device with AR and various softwares for educational use.
Honors & Awards
-
Spirit of Sussex Award: Gold
University of Sussex
-
Shine award for science and innovation
-
More activity by Piyush
Other similar profiles
Explore collaborative articles
We’re unlocking community knowledge in a new way. Experts add insights directly into each article, started with the help of AI.
Explore MoreOthers named Piyush Agrawal
-
Piyush Agrawal
-
Piyush Agrawal
Vice President, Information Security & Privacy at Genpact
-
Piyush Agrawal☁️
-
Piyush Agrawal
-
Piyush Agrawal
938 others named Piyush Agrawal are on LinkedIn
See others named Piyush Agrawal