Abhinav Rawat Email & Phone Number

Tempe, Az, Us

Phoenix, Arizona, United States

Pennsylvania, United States

I have undertaken a project aimed at differentiating between text typed freely and text typed using Deep Learning and NLP techniques. In this work, I am focussing on the creation of novel textual embeddings using hybrid keystroke-text pairs, keystroke interval timings, and keystroke hold timings for binary classification. By leveraging these hybrid embeddings, I am attempting at developing a method for accurately differentiating between the two types of text input. This research has potential applications in a variety of fields, such as cybersecurity, user authentication and plagiarism detection, where distinguishing between free text and machine-generated text can have significant implications.

New Delhi, Delhi, India

Differential Equations (M-IV)Course code - MTH204

New Delhi, Delhi, India

I created an end-to-end pipeline, which involved custom dataset generation and designing a model architecture that resulted in generating high-quality images corresponding to each line of the Hindi story. I successfully generated images corresponding to every line in a Hindi story. In doing so, I explored various image generation models, including latent diffusion models, Generative Adversarial Networks (GANs), and Discrete Variational Autoencoders (VAEs). I also studied existing English story-image generation models architectures such as StoryGANs and StoryDALLE to inform my work. This work has potential applications in storytelling, content creation, and media production.

New Delhi, Delhi, India

I assisted in the development of a multi-sensor annotation tool that is capable of creating annotations for camera images by projecting annotations from the 3D space into the image domain, as well as creating independent 2D annotations and point cloud visualizations. I also led the backend integration of models such as Yolov5 and YoloP for object detection and object segmentation, with custom JSON saving capabilities. Additionally, I integrated an Active Learning based frame selection using the CDAL algorithm to facilitate the efficient and optimized training of object-detection models. This tool has the potential to significantly improve the accuracy and efficiency of object detection and segmentation tasks especially on Indian roadway datasets, benefiting a wide range of industries such as robotics, autonomous driving, and surveillance.

Noida, Uttar Pradesh, India

We investigated the problem of student engagement in academics and explored ways to create more engaging content using video data. We employed various techniques for feature extraction, including facial landmarks, iris landmarks, and Heart Rate Variability (HRV). To ensure accurate data collection and analysis, we also developed and implemented custom annotation protocols. We compared and trained different models, such as Random Forests, Support Vector Machines (SVMs), and Naive Bayes models, to gain insights and enhance the research outcomes. The outcomes of my research have the potential to contribute to the development of more engaging academic content and improve student engagement in academics.

New Delhi, Delhi, India

I assisted in the development of the first Platform for AI and AR-based end-to-end visually empowered Inspection as a Service (IaaS). In this capacity, I worked to ensure that the platform was both efficient and effective. My role also included collecting and preprocessing data, using annotations on damaged and undamaged automobile parts to train the AI and AR models. The development of this platform has the potential to revolutionize inspection services and improve the accuracy and speed of identifying damaged parts.

Master Of Science - Ms, Data Science, Analytics And Engineering, 3.9

Bachelor Of Technology - Btech, Electrical, Electronics And Communications Engineering

Activities and Societies: Football Coordinator (2020 - 2022) Football Captain (2022 - 2023) Sankalp NGO volunteer (Jan 2022 - May 2022).