How To Decide If You Need A Machine Learning Engineer On Your Team

Here's why a Machine Learning Engineer could be the missing piece for your data team.

In modern data teams, there are a lot of roles. But first

The global machine learning (ML) market is projected to grow from $15.50 Bil in 2021 to $152.24 Bil in 2028 - www.fortunebusinessinsights.com

10x increase of ML tooling means, 🤯 growth of ML applications.

As a business leader, you have been building a world-class data team. Recently, you have been considering adding a Machine Learning Engineer (MLE). You have no idea if you really need the emerging skillset of an MLE and how to be successful in working with one.

I have been an MLE for a significant part of 2021, let me help you

1. What are the value props of MLE on your team?

ML engineers primarily bring ML/AI to your end-users

Software is rapidly reducing the effort to find a good ML model to solve your business problem. The two hard parts left are: framing the business problem into the correct ML problem and productizing the predictions. The former is done by Data Scientists and the latter by MLEs.

In other words, they build model training and serving infrastructure

2. How to know if an MLE is good or bad? Ideally, before hiring them!

Great MLEs are SWE with spikes in Distributed Systems and ML.

I recommend you look for people who are

Similar to Sr. Engineers in your product teams +

Experience with ML frameworks (Keras, PyTorch, Tensorflow ..) +

Strong Communication Skills

Lastly, it might be relevant to see if you also need them to have experience with one of the ML application domains, i.e. Recommendation Systems, NLP, Computer Vision, Optimization, or Info retrieval.

3. What should a 30/60/90 day plan look like for your first MLE hire?

Set your MLE for success, create a plan with achievable expectations

30 Days: Understand the business domain and current tools used by data scientists to solve them. Identify the gaps to build and serve ML models.

60 Days: Build version 1 (v1) of the build-deploy-monitor lifecycle for the simplest project. Educate the team and get feedback with an internal launch.

90 Days: Release v1 to users. Plan and initiate execution on scaling your work from 1 project to multiple. This can include hiring more people, vendor selection for the company, etc.

If you’ve made it all the way here, thanks for reading :) I wish you all the success in your journey to staffing and organizing MLEs.