Akshat Jaggi

Experience

• 

  Innomatics Research Labs

  Jul 2022 - Oct 2022

  Data Science Intern
• 

  GlobalShala

  Jul 2022 - Aug 2022

  Data Visualization Intern

  Python | Data Analytics | MS - Excel

• 

  Manipal Institute of Technology

  Jan 2023 - May 2023

  Undergraduate Researcher

  In this work, we investigated different techniques for out-of-distribution detection in order to enhance model performance and user perception of the models’ reliability. We chose a few publicly accessible deep learning modelsand applied them to the input data type of images. We made use of publicly available datasets of image datatype like CIFAR 100, CIFAR 10, iSUN, LSUN, SVHN where CIFAR 10 and CIFAR 100 were used as in-distribution datasets while others used as out-of-distribution datasets. Then, we showed that these models make irrational predictions on datasets that are out of distribution. We were interested in showing the great accuracy with which out-of-distribution data may be detected using approaches like Mahalanobis distance, Softmax score, ODIN, and Energy score-based OOD detection methods. We next converted the OOD score into a confidence score, which is understandable by humans, in order to examine its impact on how users engage with ML applications. With this work, we hope to show that data shift is a crucial knowledge for ML-based high-stakes apps like medicine and health diagnosis. Show less

• 

  GSK

  Aug 2023 - now

  Technical Associate

  Python, Powershell, Terraform, Pyspark, Azure Databricks, Data Management on Azure, Azure SQL Server Management Studio, Azure Devops, MS - Excel