Hi! 😃

My name is Debjyoti Mondal. I’m a Research and Development Engineer at Samsung.
And I like a lot of things.

</img>

Work Experience

• Samsung Research Institute, Bangalore
       Research and Development Engineer ~ June 2022 - Present
       I work in the Natural Language Understanding team. Helping models reason and maintaining a few key modules in the Bixby pipeline.
       How are we doing? Can’t say much, but check this out.

     Student Trainee ~ May - July 2021 📜
     Worked on the interpretability of Language Models for Intent Classification and Slot Tagging tasks.
     Used Language Interpretability Tool (LIT) and Local Interpretable Model-agnostic Explanations (LIME).

• NxTechWorks Consulting Pvt. Ltd., Pune
       Machine Learning Intern ~ Dec 2020 - Jan 2021 📜
       I developed a generic object detection and localization framework.
       And also some OCR apps for template based info extraction.
       Hehe, this was fun - implemented deskew and denoise for scans using Hough transform and Thresholding.

Education

• Indian Institute of Technology (ISM), Dhanbad
       Bachelor of Technology in Electronics and Communication Engineering ~ July 2018 - May 2022

Publications

• KAM-CoT: Knowledge Augmented Multimodal Chain-of-Thoughts Reasoning 📄
       - We introduce knowledge graphs as a new modality, keeping the graph structure intact.
       - With 2 graph convolutional layers as the KGEncoder, our method gains a deeper contextual udnerstanding of text,
       therefore reducing hallucinations and enhancing the quality of answers.
  

     Association for the Advancement of Artificial Intelligence (AAAI), 2024

• Seg-HGNN: Unsupervised and Light-Weight Image Segmentation using Hyperbolic Graph Neural Networks 📄
       - We exploit the properties of Hyperbolic Spaces to do segmentation tasks in very low dimensions.
       - With the extracted image features, we build a graph in the Lorentz space,
       and perform unsupervised clustering to get semantically similar regions.

     The British Machine Vision Conference (BMVC), 2024

Some cool projects

• Modeling Chaotic Epidemic Models on FPGAs - Bachelor’s Thesis
       This was a general procedure to implement Chaotic Models.
       I used Euler’s Forward method to discretize continuous-time differential equations,
       then generated RTL Schematics and compared the resources used.
       Ahaa! And using this, I studied period-doubling bifucations & got to the Figenbaum constant
       with the logistic map and the Genesio-Tesi attractor. Be sure to watch this. It’s pretty.
  
  
• Neural Style Transfer
       Implemented the paper, Artistic Style Transfer, by Gatys Leon.
  
  
• Plant Pathology - FGVC7 Kaggle Competition
       Developed classification models to identify foliar diseases in apple trees. Used an ensemble of ResNet50 and EfficientNetB7.
  
  
• Deep Dream
       Implementation of the DeepDream algorithm. Extracts features from InceptionV3, and using gradient ascent over iterations, enhances trippy patterns in the input image.