CGPA: 9.0/10.0

Scaler is an online transformative upskilling platform for working tech professionals.

Roles/Responsibilities:
• Setting up problems/quiz for Scaler Topics on Keras & PyTorch.
• Working to create an outline for technical content for Keras.
• Writing different technical content as a blog post for Keras.

Roles/Responsibilities
• Implemented MobileViT in TensorFlow and Keras
• Used MobileViT's official Pytorch weights to port from Pytorch to TensorFlow
• Created two notebooks for tutorial purposes i.e Off-the-shelf classification and Fine-tuned notebook.
• Publish all the models in TensorFlow Hub.

TensorFlow Hub is the main TensorFlow model repository with thousands of pre-trained models with documentation, sample code and readily available to use or fine-tune. The idea behind the project is to develop new State-of-the-Art models like CoAtNet and MobileViT and publish the pre-trained models on TensorFlow Hub using the ImageNet1k dataset. CoAtNet is considered the State-of-the-Art model for Image Classification whereas MobileViT enables us to make inferences in edge devices while using Transformers.

Mentors: Luis Gustavo Martins and Sayak Paul

Roles/Responsibility:

• Responsible for creating Computer Vision Services to enhance the product.
• Worked on on-device Deep Learning with TensorFlow Lite as well deploying Machine Learning Services in
production with AWS Sagemaker.
• Trained models on Google Cloud Platform, reproduced SOTA models from research. papers.

Blend is a mobile 1st AI-driven graphic design and marketing app for entrepreneurs that removes decision-making to create professional-grade product visuals, social graphics, and product descriptions.
Just upload your product photo and Blend does the rest.

Roles and Responsibilities:

• Understand the Tensorflow Lite Task Library.
• Implemented CameraX and remove the usage of fragments with the existing Camera2 and Camera API in Object Detection App
• Implement Support Library with TensorFlow Lite Interpreter.
• Implemented Data Binding
• Implemented the Image to BitMap conversion and modify Support Library and Task Library
• Modified the Test Code in Android
• Adding a Bounding Box Function in tflite-support Library.

Role/Responsibilities
• Build many Computer Vision Models as per customer requirement. Optimised the models with various optimisation technique like Quantisation, Weight Pruning and Weight Clustering.
• Played a major role for developing Ikshana AI. Worked with TensorFlow Object Detection API, TensorFlow JS and TensorFlow Lite to build the model and deploy it in edge devices.
• Worked with big companies like Mahindra and PropDial.

• Identified Software and DB Architecture based as per customer requirements, performed version control and refining product.
• Created the backend using Nodejs, Express, MongoDB, MySQL and maintained Digital Ocean Server with Droplet and Storage bucket.

Roles/Responsibilities
• Implemented MobileViT in TensorFlow and Keras
• Used MobileViT's official Pytorch weights to port from Pytorch to TensorFlow
• Created two notebooks for tutorial purposes i.e Off-the-shelf classification and Fine-tuned notebook.
• Publish all the models in TensorFlow Hub.

American Sign Language Detection is a deep learning end to end project where we can detect American sign Language. It handles upto 29 classes. Used MobileNetV2 to train the images. It is deployed in smartphone using TF-Lite.

The idea behind the project is to improve the sample apps of Computer Vision which uses TensorFlow Lite Task Library as well as TensorFlow Support Library. The main objective is to implement CameraX and update the app so that the developers from the community find it easy to integrate Machine Learning with Android Apps.

This project demonstrates the use of TensorFlow Object Detection API to automatically detect Red Blood Cells (RBCs), White Blood Cells (WBCs), and Platelets in each image taken via microscopic image readings.

ConvMixer, an extremely simple model that is similar in spirit to the ViT and the even-more-basic MLP-Mixer in that it operates directly on patches as input, separates the mixing of spatial and channel dimensions, and maintains equal size and resolution throughout the network. In contrast, however, the ConvMixer uses only standard convolutions to achieve the mixing steps. Despite its simplicity, we show that the ConvMixer outperforms the ViT, MLP-Mixer, and some of their variants for similar parameter counts and data set sizes, in addition to outperforming classical vision models such as the ResNet.

A deep learning model is build to detect malaria of the patients with the cell images of the patients.

CodeChef KIIT Chapter, an initiative for the students who are enthusiastic about programming.

Roles/Responsibilities
• Manage the whole team.
• Organise sessions and events.
• Guide and Mentor various students.

A group of dedicated students and faculties who are keenly focused on working towards the betterment of society through technology believing that small steps gradually contribute to bigger and better changes.

Even though we are a research lab, we harbour all domains, technical and non-technical for a complete in-house holistic appraoch towards everything.