My research interests include Machine Learning, Deep Learning, Reinforcement Learning, Computer Vision, Cloud Computing, Database Systems and Distributed Systems. My passion lies in developing multimodal solutions using domain knowledge and applied machine learning. All my published work is available through Google Scholar and Research Gate.
This paper studies DeliverAI’s performance on the multiple objectives of the routing problem and the decentralized execution of the RL agents. Simulations on real-world data from Chicago show a 65% superior performance on fleet management, 15% reduction fleet size, and 15% reduction in operational costs in comparison to SOTA baselines. Contributions also include an open-source delivery dataset and simulator.
SLEM is the first model to align spectrogram embeddings from audio with language embeddings from lyrics. I leverages this multimodal integration to improve genre classification accuracy. This paper presents an extensive study on the linear combination of the two modalities using state-of-the-art deep learning models and tests it on the song previews from Spotify API.
This work was also presented at the The IEEE World Congress on Computational Intelligence (IEEE WCCI 2024) - See Presentation
This work proposes the first "multi-hop ride-sharing" architecture for food deliveries using a multi agent Q-learning based algorithm. It models the traditional routing problem as a Markov Decision Process. It also solves pressing challenges of an exploding action space by developing a novel clustering framework to reduce complex city network into an overlay network of hotspots.
I have peer-reviewed for the following Conferences and Journals:
IEEE Artificial Intelligence Review (AIR)
European Conference on Machine Learning (ECML)
IEEE International Joint Conference on Neural Networks (IJCNN)
Springer Real-Time Image Processing
Springer Signal, Image and Video Processing
Springer Scientific Reports
Journal of Computational Methods in Sciences and Engineering
Papers Reviewed
