Vijay Anand Raghava Kanakagiri

Experience

• 

  Plumsoft Pvt Limited

  Jun 2017 - Jul 2017

  Android Developer Intern

  I designed app which the company uses for geo tagging for their products and I also offed support for their websites through app and also provided integral support for their current websites on apps.

• 

  National Informatics Center

  May 2018 - Jul 2018

  Web Development Intern

  Designed and improved the API calls from backend by following MVC pattern using ASP.Net for Digitalizing Land Records.Created the database backend framework for storing records related to land and their respective owners (size, purchase history, current market price, etc.) belonging to a given state.

• 

  Blaize

  Jan 2019 - Jul 2022

  I was responsible for finding scalable, robust solutions to problems through literature survey, ideation, and conduction of experiments. Also responsible for code stabilization, generalization, and release of the solution proposed.Designed and developed Auto Mixed Precision and Quantization algorithms which retained 98% floating-point accuracy for existing low bit Image Classification, Detection, and Segmentation models while also boosting its performance i.e. (FPS). Also added pruning, sparsity analysis support, and later extended this project by adding Quantization Aware Training to handle imbalanced data distribution of layers.Worked on developing framework and Neural Network Support for the ‘AI Studio’ product which delivers AI-driven DataOps, DevOps, MLOps and is the key revenue generator for Blaize. As a research engineer in field of artificial intelligence and neural network, some of my work dealt with implementing recent and advanced neural networks and porting them into company specific hardware. Frameworks like Pytorch, Tensorflow, Keras, Onnx, Openvx were used to train the neural networks and deploy them on hardware, depending on projects which gave a good grasp on these various frameworks.Worked on developing ISP Pipeline which processes images in different stages from Raw Image format to RGB/YUV Format. Show less Performed Spatial and network decomposition on convolution nodes effectively improving its performance on parallel compute by a factor of 125% - 350% on many recent Object Detection and Image Classification models.Trained a deep neural network in fixed point domain with reduction in bit width and also introduction integer batch normalization. It also included training the Neural Network specific to hardware and also training them in fixed point representation as it is required to achieve high optimization. Show less

  • AI Research Engineer 2 (Computer Vision & Artificial Intelligence)

    Jun 2019 - Jul 2022

  • AI Research Intern

    Jan 2019 - May 2019
• 

  Texas A&M University

  Jan 2023 - Jun 2023

  Teaching Assistant

  Graduate Teaching Assistant for CSCE 120/121 - Introduction to Program Design and Concepts

• 

  Halliburton

  May 2023 - Aug 2023

  Software Development Intern
• 

  Texas A&M University

  Aug 2023 - now

  Teaching Assistant

  Graduate Teaching Assistant for CSCE 482 - Senior Capstone Design