Baking Machine image with confidence

Devesh Bajaj (~devesh50)


With machine learning increasingly becoming an engineering problem -not all ML applications are compatible with docker, and being an event-driven company adds the cherry on the top. Most of the standard approaches like docker, rtk fails to solve the problem like heavy model deployment, multiprocessor apps.

At Episource, we faced a similar challenge of deploying 100s of instances and configuring these instances at boot time with a wide variety of tools and libraries to support the inference models. This installation of ML model and configuration them at runtime comes with the cost of both time and money. Any failed bootstrapped instance can make that instance obsolete, which leads to affecting our SLA.

Avoid any runtime installation and network call for installation, we have built the solution around machine image and its interaction with our CICD pipeline.

This talk revolves around the best practice to build Machine images to avoid runtime crashing and how to automate the same workflow in release CICD and reduce the developer overhead. This talk talks don't focus on the specific tool but on the approach to build a sustainable and strong MI building CICD pipeline.


  1. A quick primer on problems related to Implement a CICD on Non-Dockerise application (~ 5 mins)
  2. Introduction about Episource and speaker (~3 mins)
  3. Introduction to Machine Image and its multiple use cases (~2 mins)
  4. Tools and practice associated with Machine images (~5 mins)
  5. Problem with building image in CI server (~5 mins)
  6. Observability and testing pipeline of image building (~5 mins)
  7. QnA (~5 mins)

Slides for the talk

(Note:- Slides are subject to vary a little at the time of conference )


  1. Basics of github (here)
  2. Little to no knowledge of CICD (here)
  3. Basic knowledge of machine Images (here)

Speaker Info:

Devesh is a Data Engineer with Episource LLC, with close to 2 years of experience.

  • With primarily involved in building microservices that perform Machine Learning Inference along with the development and deployment of end-to-end full-stack applications.

  • With the keen passion for the development of IoT based application and Won 2nd​ ​ prize​ ​ in​ ​ ​ CanSat​(2017) ​ Satellite​ design ​ competition​ ​ organized​ ​ By​ ​ American Astronautical​ ​ Society​ ​ (AAS)​ ​ and​ ​ Recognised​ ​ by​ ​ NASA.

Speaker Links:

You can find Devesh at

LinkedIn :-

Twitter :-

Medium : -

Github :-

Section: Others
Type: Talks
Target Audience: Beginner
Last Updated: