About

Karan Jariwala is a Machine Learning Engineer II at Amazon, specializing in Explainable AI to detect pre-training bias, post-training bias, and Model explainability of black box and white box model. He has 8.44 years of relevant experience in the field of Machine Learning, Deep Learning, and Computer Vision. Karan has worked on building and enhancing the Computer Vision Model Explainability capabilities for Amazon SageMaker Clarify. He is also working on enhancing the Model Explainability for Tabular Modal type. He has a good knowledge of Model explainability methods such as Grad-CAM, Integrated Gradients, Layer-wise Relevance Propagation, Taylor Decomposition, Deep LiFT, etc. Before working on Explainable AI, Karan worked on building the SageMaker Distributed DataParallel library and did research and optimization of Computer Vision models such as Mask R-CNN, Faster R-CNN, etc. to run best on AWS. He was also a part of Apache MXNet backend for Keras deep learning library. Karan has worked as a Deep Learning Engineer I at Amazon, where he worked on the SageMaker Distributed Data Parallel or Herring, a distributed training library using PyTorch and TensorFlow framework that is optimized for large scale distributed training at AWS. He also worked on integrating Horovod using Apache MXNet on AWS Deep Learning Containers(DLC), Deep Learning AMI(DLAMI), and SageMaker. He has also mentored the development of Horovod Integration on SageMaker. Karan's work on optimizing the training time of Mask R-CNN model using Apache MXNet from three hours to 25 minutes on 24 Amazon P3dn.24xlarge EC2 instances during AWS re:Invent 2019 is noteworthy. Karan has also worked as a Graduate Research Assistant at Rochester Institute of Technology, where he created a new positional dataset from MS-COCO, Visual Genome, and Visual7w with sentence about the image. And applied the Faster-RCNN model on the Image and Skip-thought vector with GRU on sentence/question to generate multiple bounding box as the answer. Karan has a Master’s Degree in Computer Science from Rochester Institute of Technology, a Post-Graduate Diploma in Embedded Systems & Design from Centre for Development of Advanced Computing (C-DAC), Pune, and a Bachelor of Engineering (B.E.) in Electrical, Electronics and Communications Engineering from Sarvajanik College of Engineering & Technology. His Tech Stack