Hey there! 
I’m Rugved, a passionate Computer Engineer, Deep Learning enthusiast, and a New York University graduate passionate about building high-performance, scalable systems. I’m currently working as a Software Development Engineer at Annapurna Labs (AWS), where I help manage the ML Accelerator Fleet.
With over two years in DevOps at Oracle and a Bachelor’s degree in Electronics Engineering from the University of Mumbai, my journey in technology has been dynamic and invigorating - spanning across cloud infrastructure, machine learning, system optimization and hardware-software co-design.
I’m excited to share my experiences, insights, and explorations here!
I’ve previously taught…
Here are a few things I have worked on…
-
Multimodal Sentiment Analysis - This project explores different methodologies, focusing on fusion techniques that integrate information from three modalities (audio, language, vision) to enhance sentiment analysis performance, and the various architectures that employ them. [Read More]
-
Continual Learning for Autonomous Vehicles - Applying the continual learning framework to Autonomous Vehicle setting: predicting steering angles based on the captured images. Experimented with various approaches - Elastic Weight Consolidation, Experience Replay and Temporal Consistency Regularization. [Read more]
-
Vector Processor Simulator - A Python-based Vector Processor (VMIPS) functional and performance simulator evaluated on dot product, fully connected layer, and convolutional layer test bench. [Read more]
-
ResNets - Explored ResNet Models trained to classify images on CIFAR-10 dataset. Finetuned hyperparameters, experimented with ResNet depths and strategies like dropout layers and learning rate schedulers achieving a test accuracy of 95.37%. [Read more]
-
Semantic Image Segmentation for Autonomous Vehicles - This project utilizes semantic image segmentation to enhance autonomous vehicle perception, accurately identifying and segmenting critical elements like drivable surfaces. By leveraging a U-Net model trained on the CARLA dataset, the project achieves an impressive 96% accuracy [Read more]
