Hi there!

I am Mayukh Bhattacharyya. I am a Software Engineer with over 4 years of experience at companies like Snap Inc, Samsung and Applied Materials with an experience of projects in diverse domains ranging from Recommender Systems, Computer Vision and recently Large Language Models. My expertise and interests lie at the juncture of Software Development and Machine Learning. I am passionate about building products that people would love. The satisfaction of creating something useful keeps me going!

I am somewhat active on the ML community of Kaggle, where I tend to share my works. I am on Master Level in the Notebooks track and hold above 99 %ile ranks in both the Competitions and Notebooks tracks.

I wrote reco, a RecSys library, in the past when there were not good enough RecSys packages. With a Cython based core (same as what Pandas uses), it was faster than most other contemporary libraries and found some easy adoption thanks to its Scikit-Learn style design.

I sometimes write a few articles on medium. This may be an interesting read where I analysed our environmental impacts at the onset of COVID-19.

Publications

DeCAtt: Efficient Vision Transformers with Decorrelated Attention Heads. Mayukh Bhattacharyya and Sayan Nag.
Efficient Deep Learning for Computer Vision Workshop, CVPR 2023

Deciphering Environmental Air Pollution with Large Scale City Data. Mayukh Bhattacharyya, Sayan Nag and Udita Ghosh. (Best Paper in AI for Good Track)
International Joint Conference on Artificial Intelligence (IJCAI 2022) [code] [pdf]

SERF: Towards better training of deep neural networks using log-Softplus ERror activation Function. Sayan Nag and Mayukh Bhattacharyya
IEEE/CVF Winter Conference on Applications of Computer Vision (WACV 2023) [pdf]

Hybrid Style Siamese Network: Incorporating style loss in complementary apparels retrieval. Mayukh Bhattacharyya and Sayan Nag.
Workshop on Computer Vision for Fashion, Art and Design, CVPR 2020 [code] [pdf] [video]