Saurav Maheshkar

ML Student
⚖️ 🧐
Hi! My name is Saurav Maheshkar and I'm a Senior High School Student. I love learning about Machine Learning and Artificial Intelligence. I'm presently working on Attention and Self-Supervised Representation Learning.

🩺 X Ray Image Classification

⚖️ 🧐
Submitted as a part of Deepnote publishing competition

Link to the Published Notebook

I go over the entire pipeline of creating a Binary Image Classifier using Tensorflow. I cover, all aspects of the pipeline such as experimenting with different network architectures and comparing metrics, pruning and quantization of the model for faster inference and finally two methods of deploying such models. We'll train a Convolutional Neural Network to recognize chest X-rays of people with pneumonia and then create an interactive web application.

For this project I also used Weights and Biases for experiment tracking and their artifacts feature for storing the model weights. Link to the wandb project page.

Please ⭐️ the Github Repository, if you like the project.
Like Comment

Named Entity Recognition using Reformer

⚖️ 🧐
Reformer, which was proposed in the paper Reformer: The Efficient Transformer by Google Research, essentially addresses some efficiency constraints of the Transformer model and proposes an improved version of the Transformer that implements Locality Sensitive Hashing (LSH) and Reversible Layers to make the model much more efficient.

I’ve made a kaggle kernel implementing the Reformer Model for Named Entity Recognition(NER). Please review the notebook and tell suggestions.
Like Comment


⚖️ 🧐
I’m a senior high school student from New Delhi, India and I want to pursue deep learning in my higher education. I spent the last 2 years studying Machine Learning having completed multiple specialisations from, Google, IBM, Imperial College London and many other courses (ALL CERTIFICATES ON MY WEBSITE). Before that I’ve worked as a Front End Developer with 6 months of internship experience in a leading educational startup. I also have a tech firm which I started with a couple of friends and we take industry projects for App and Web development. I’ve also worked a lot with IoT and Robotics having won multiple hackathon’s and representing my projects in the National Science Fair consecutively for 2 years. Two of my projects on Fake News Classification and Regression Analysis of Primary Biliary Cirrhosis are currently being mentored by Intel under the AI 4 Youth program.

I hope to pursue my Bachelors ( Hopefully with integrated masters ) from the United Kingdom. I’ve applied and already gotten an offer from the university of Birmingham, hoping to get into University of Edinburgh and Oxford as well. (Fingers Crossed).

I’m currently working on NLP with attention to the Reformer Architecture and Self-Supervised Representation Learning.

PS: Please check out this kaggle kernel, where I’ve tried to use the Reformer model for NER, using the Trax library. It’s the only kernel on kaggle using the Reformer model for any sort of analysis. Thank You. Excited to connect with such experience industry leaders.
Like Comment