Heroku is a cloud platform as a service (PaaS); I work on the Heroku Identity team to design, develop and maintain web services (APIs & internal tooling).
•
Developed an update for the Heroku API to emit events and log information in the audit-logs when an Enterprise Account member is created, updated or destroyed.
Worked on a galaxy morphology classification task with deep convolutional neural networks, particularly a variant of Residual Networks (ResNets). The variant is applied to a sample of 28790 images from the Galaxy Zoo 2 dataset, to classify galaxies based on their morphology followed in this paper (link). The primary objective of this task is to run the task using a workflow management system called Pegasus that automates scientific computational work as portable workflows.
•
Designed the deep convolutional neural network to use multiple threads in a distributed computing environment to run most of its executables in parallel batch processing. Used Docker to containerize this workflow, and along with the Pegasus API to run on a national, distributed computing partnership for data-intensive research The Open Science Grid.
•
Designed and developed a traditional VGG Convolutional Neural Network using transfer learning to classify Cat and Dog images to perform binary classification. The model was implemented in Keras using the Pegasus API and containerizing the script with Docker .