• Overview
  • Technical Features
  • Takeaways
  • Doordash MLOps

    Model Profiling using Pytorch

    PyTorch | Python

    Link: Blog

    Overview

    Introduction

    DoorDash is an online food delivery and ordering service. DoorDash wants us to build an ML environment where models can be tested and run by ML Engineers to ensure they work properly without risking production!

    Our DoorDash team built a testing platform from scratch to empower engineers to comprehensively evaluate ML models. The platform was rigorously tested for invalid outputs and out of range values to indicate model drift. We also caught issues in production early using predictive insights and evaluation of systemic behavior.

    Solution

    DoorDash is an online food delivery and ordering service. DoorDash wants us to build an ML environment where models can be tested and run by ML Engineers to ensure they work properly without risking production!

    Our DoorDash team built a testing platform from scratch to empower engineers to comprehensively evaluate ML models. The platform was rigorously tested for invalid outputs and out of range values to indicate model drift. We also caught issues in production early using predictive insights and evaluation of systemic behavior.

    Deliverable

    DoorDash is an online food delivery and ordering service. DoorDash wants us to build an ML environment where models can be tested and run by ML Engineers to ensure they work properly without risking production!

    Our DoorDash team built a testing platform from scratch to empower engineers to comprehensively evaluate ML models. The platform was rigorously tested for invalid outputs and out of range values to indicate model drift. We also caught issues in production early using predictive insights and evaluation of systemic behavior.

    Technical Features

    Synthetic Data Generation

    DoorDash is an online food delivery and ordering service. DoorDash wants us to build an ML environment where models can be tested and run by ML Engineers to ensure they work properly without risking production!

    Our DoorDash team built a testing platform from scratch to empower engineers to comprehensively evaluate ML models. The platform was rigorously tested for invalid outputs and out of range values to indicate model drift. We also caught issues in production early using predictive insights and evaluation of systemic behavior.

    none

    Model Profiling

    DoorDash is an online food delivery and ordering service. DoorDash wants us to build an ML environment where models can be tested and run by ML Engineers to ensure they work properly without risking production!

    Our DoorDash team built a testing platform from scratch to empower engineers to comprehensively evaluate ML models. The platform was rigorously tested for invalid outputs and out of range values to indicate model drift. We also caught issues in production early using predictive insights and evaluation of systemic behavior.

    none

    Model Drift Detection

    DoorDash is an online food delivery and ordering service. DoorDash wants us to build an ML environment where models can be tested and run by ML Engineers to ensure they work properly without risking production!

    Our DoorDash team built a testing platform from scratch to empower engineers to comprehensively evaluate ML models. The platform was rigorously tested for invalid outputs and out of range values to indicate model drift. We also caught issues in production early using predictive insights and evaluation of systemic behavior.

    none

    Takeaways...

    DoorDash is an online food delivery and ordering service. DoorDash wants us to build an ML environment where models can be tested and run by ML Engineers to ensure they work properly without risking production!

    Our DoorDash team built a testing platform from scratch to empower engineers to comprehensively evaluate ML models. The platform was rigorously tested for invalid outputs and out of range values to indicate model drift. We also caught issues in production early using predictive insights and evaluation of systemic behavior.