rhitviksinha Goto Github PK

• Undergraduate Level Classes:
• CS10001: Programming & Data Structures (8/10)
• MA20101: Transform Calculus (8/10)
• MA20106: Probability & Stochastic Processes (9/10)
• CS31702: Computer Architecture & Operating Systems (8/10)
• CS40032: Principles of Programming Languages (8/10)
• AI42001: Machine Learning (9/10)
• Graduate Level Classes:
• AI60004: Big Data Processing (9/10)
• AI61002: Deep Learning (9/10)
• CD66001: Accelerated Data Science (10/10)
• AI61005: Artificial Intelligence (8/10)
• CS61061: Data Analytics (9/10) (9/10)
• CS66010: Security & Privacy in Online Social Networks (8/10)

• CSCI-GA 1180: Fundamental Algorithms (confirmed)
• CSCI-GA 2565: Machine Learning (confirmed)
• CSCI-GA 2271: Computer Vision (subject to pre-requisite waiver)
• DS-GA 1014: Optimization and Computational Linear Algebra (subject to DS dept. approval)
• CSCI-GA 2110: Programming Languages (can also be taken next semester)

Working on Transformational AI projects in the Health Sector, as part of Ministry of Electronics and Information Technology's Digital India Internship Scheme 2021 (2nd Cycle). Responsible for the landscaping and the subsequent review of the ecosystem and policies operating at the intersection of AI and Healthcare in India.

• Working under the guidance of Prof. Adway Mitra on Extreme Weather Prediction and Climate Modeling using Deep Learning.
• Dataset Used: NCAR CESM-LENS (stored on AWS S3)
• Python-based implementation, PyTorch used as Deep Learning framework of choice to implement “Dynamic Routing Between Capsules” (2017, Hinton et al.) from scratch.
• Recreate findings from “Analog Forecasting of Extreme-Causing Weather Patterns Using Deep Learning” (2020, Chattopadhyay et al.)

Apply pre-trained Deep Learning models for the Human Pose Estimation task. Use extracted Pose Information to suggest improvements in a Cricket (Sport) Learner's technique, preventing inefficiency and injuries.

• Python based implementation, with Deep Learning framework of choice being PyTorch.
• Torchvision's KeypointRCNN class finetuned for best accuracy used for the Human Pose Estimation (HPE) task. Used to generate time series pose data from video clips.
• Recurrent Neural Net used for maximizing accuracy at detection of potential errors in technique from gathered time series pose data.
• Company has since pivoted to a metaverse oriented application, and had it's sports division shut down.

• Automate last mile delivery through self-driving rovers that operate on sidewalks.
• Setting up of cloud-based training and deployment pipelines for Deep Learning models utilizing AWS services, including but not limited to: S3, EC2, SageMaker
• Pipelines for the following frameworks were setup: TensorFlow, PyTorch