by Aabir Abubaker Kar
~1 min read

Categories

  • posts
  • projects
  • Projects

Tags

  • deep learning robotics data science

Quick summary:

  • What I did:
    • Co-developed a PyTorch LSTM trained on AWS EC2 to detect spatial obstacles in real-time with 84% true positive rate
    • Coded and deployed a 90% data verification ETL pipeline for flight logs using ARIMA time-series forecasting and outlier detection
    • Co-designed and presented 6 Business Intelligence dashboards to a 300-person audience at company-wide weekly meetings
  • Things I learnt:
    • How to productionize machine-learning pipelines - from simple ARIMA models to complex deep neural LSTMs
    • Business Intelligence thinking - how to help business strategy by metricizing performance
    • How to work with cloud and related technologies like Amazon Web Service (EC2, SageMaker, S3), Sisense dashboards, Docker, proprietary package management software
  • When was this?
    • June 2020 through September 2020

I spent summer of 2020 at Zipline. It was entirely virtual due to the pandemic, but still a rollercoaster of an experience.

Shoutout to my wonderful colleagues at Zipline who taught me so much - Matt Fay, Anne DeGraaf, Hugo Carr, and so many others.