Stay up to date
Subscribe to our newsletter
Machine learning engineering is not an entry-level job. To become a machine learning engineer, you need experience in data science and data engineering, as well as a higher education degree in a relevant domain. Luckily, for professionals who are interested in entering or transitioning into the world of machine learning engineering, there are many training programs, online and offline courses, and free material on the web to help them in this transition.
Throughout this guide, machine learning engineering and recruitment professionals at Xomnia provide an overview of the role of a machine learning engineer, the steps to begin and maximize your career in machine learning engineering, as well as ways to stand out in the job market.
What is the role of a machine learning engineer?
Working with models from A to Z
A machine learning engineer’s role is focused on the deployment of models and bringing them into production to achieve actual value, according to Xomnia’s Machine Learning Engineer Pelle Wage. MLEs are the ones focusing on the “A-Z aspect” of starting with data and ending with a product that can be used.
“You are expected to use your analytical and data science capabilities to not just make a model or an analysis, but to also turn that into something that can be used continuously,” explains our MLE Hella Haanstra.
Critical thinking and communication skills
Being a successful machine learning engineer, particularly at consulting firms, requires more than purely technical skills. Transferable skills, such as critical thinking and communication skills, are important for an MLE.
“You really need to be a critical thinker to address the challenges that face putting a model in production,” says our Machine Learning Engineer Martijn Beeks. “You also need to know and understand the business domain you work for, and align this with the technical side of projects..”
MLE-specific technical skills
On a more technical level, there are skills that are unique to a machine learning engineer. According to our Lead Data Engineer Josko de Boer, those include, for instance, MLOPs setups and DevOps principles of machine learning.
“There are numerous concerns that require machine learning engineering that a data engineer can’t do, such as more massive deployments of ML, involving concerns on scaling, latency, IOT devices, managing the (re)training, deployment, etc” explains Josko.
Ability to help with DE & DS principles
Last but not least, a machine learning engineer is expected to be familiar with the principle of data science and data engineering.
“Sometimes, a machine learning engineer focusses on hardcore ML topics. But there are projects in which they find themselves switching between the two roles, from working a bit on data science, to cleaning code and putting pipelines in production, before being able to work on ML concerns” adds Josko.
What qualifications do I need to be a machine learning engineer?
A relevant higher degree
At least in the Netherlands, most of those applying to work as machine learning engineers already have a masters degree under their belt, which are usually in very specific domains, such as AI, data science, and econometrics. Individuals with data engineering skills (which are taught in some computer science programs) are particularly appealing to recruiters too, as they are more likely to be able to properly structure data pipelines.
“As recruiters, we’re facing a shortage in data engineering talents because very few university programs teach data engineer skills such as deploying a model in the cloud, which are things that are usually learned on the job,” explains our Campus Recruiter Daniela Alvaran.
“Despite this trend, there still seems to be a particular interest in graduates who come from backgrounds such as computer science, as they are more likely to be able to productionalize and deploy APIs”, adds Daniela.
Ability to work with data, models and clouds
More specifically, an MLEs qualifications must include familiarity with programming in Python, and the ability towrite code, implement models, and work with data. Moreover, being familiar with cloud computing and architecture is becoming increasingly important, as the number of companies moving their work to the cloud continues to grow.
Our MLEs unanimously agree that communication skills are crucial for a successful machine learning engineer.
“It is really important to be able to translate between the technical part and your clients’ needs, because if you cannot check with your client that the product you’re making is something that they need, can use or understand, you will have wasted everybody’s time and money,” explains Hella.
“As an MLE, you deal with a wider variety of stakeholders, such as business stakeholders, data scientists who created the model, and data engineers who need to incorporate it into a data pipeline. You need to be comfortable with communicating your MLE knowledge clearly,” adds Josko.
Pelle adds being social to the qualities of a successful MLE. “Be open to learning from others and asking questions, because no one can know everything just by Googling it.” Xomnia’s Machine Learning Development Program offers a unique opportunity in this regard, because it puts its members within a network of experienced data professionals.
“When you’re part of our MLE Development Program, you can ask questions, get up to date with the latest state-of-the-art innovations that change every day, engage in peer-to-peer learning and get each other excited,” says Daniela.
Do I need a degree to become a machine learning engineer?
Our interviewed machine learning engineers unanimously agreed that the answer to this question is NO. However, they all acknowledged that the lack of a degree related to MLE is going to be a big obstacle in the way of an applicant getting asked to do a job interview.
According to Josko, this is because the MLE field is still too young for recruiters and companies to consider applicants without a masters degree. But when they start doing that and it bears fruit, he thinks that they might even be open to considering applicants without a bachelor’s. This remains an early speculation at this stage.
For Hella, whose bachelors degree focused on child education, what matters the most to become a machine learning engineer is being able to prove your experience in MLE and showcase how you have worked in projects.
“MLE comes from experience, regardless of what you learn in university, but for those who choose to go down the academic path, I think a degree in AI, econometrics, data science or anything statistical will help you know what you’re building as an MLE.”
“Coming from a background in economics, I felt a bit of a disadvantage in the job application process, as I needed to prove that I can do the job,” said Pelle, “but I was able to show my programming skills and mathematical background in the interview process, which helped me get the job.”
According to Josko, a degree is “the proof of having done the basics”, but this shouldn’t come in the place of showing that one has applied those basics. “You can grow your experience by improving your Python skills, and doing Kaggle competitions and FruitPunch challenges to get familiar with thinking with data.”
Xomnia offers a Data and ML Engineering Program that is designed to give recent graduates a DE/MLE training on a broad basis. It helps them determine what they like to do, the tools they prefer, and where to go from there. Through this program, we familiarize our MLEs in-depth about topics that are relevant or that we think will become relevant in the near future.
Can I become a self taught machine learning engineer?
Teaching yourself to become a machine learning engineer cannot take you too far. According to our experts, this is because applied experience is what constitutes a capable machine learning engineer, and not just what they know.
“My answer to this question would be no, because to become an MLE, you need the experience of working with a client, and implementation is never isolated from that because a big part of what we do involves putting something in production to be used,” explains Hella.
“Our work is an iterative process with the business, where you check in with your clients and discuss what you’re building and whether it is what they need, and this cannot be achieved when you’re working at home by yourself,” Hella adds.
For both Hella and Pelle, the transition to MLE came after working as data scientists for a while to get experience with data. They combined that with self-teaching. Moreover, Pelle joined Xomnia’s Machine Learning Development Program to enable himself to expedite his transition into MLE.
Our experts give the following recommendations for those looking to grow their MLE skills on their own time:
Andrew Ng's courses: “In Coursera, I recommend the machine learning engineering course by Andrew Ng.I am also following Ng’s MLOps course, called Machine Learning in production,” says Hella.
Coursera's Associate Cloud Developer degree: Josko recommends the Associate Cloud Developer degreefrom Coursera, and benefitting from the free trials that vendors of different cloud services offer. He warns, however, that online courses and free trials alone do not give the hands-on experience an MLE needs to have, but merely aid in that.
Learning from others: It is important to also benefit from the experience of others in the field, rather than trying to know everything on your own. “The most important thing is also learning from others; for instance, during Kaggle competitions, you can see the work of others and how they approach a problem, which is a valuable source to learn,” says Martijn.
Reading blogs: “Make use of data science and blogs to know more about what is going on in real life, and the ways to tackle a client’s problem in scenarios when there is no data or not enough data, and when clients want an easily explainable model rather than a fancy MLE model,” says Hella.
How do I start my career in machine learning from scratch?
As mentioned in the previous answer, self-study is not enough to become a machine learning engineer. This is because mastering the ins and outs of everything related to MLE needs to go hand in hand with applied experience in deploying models that solve real life problems.
This hands-on opportunity is offered, for instance, to young data & AI professionals who join Xomnia’s Data and ML Engineering Program. However, for those who need to learn the absolute basics, we recommend boot camps such as CODAM, which teach how to code from scratch, or Data Camps to learn the basics of Python from scratch.
“The first thing you need to become a machine learning engineer is learn how to program, typically on Python, because it is the language necessary to be able to build an API or work with docker or kubernetes,” says Hella. “Next, you need to move towards statistics to have a base and know how to model.”
“Just like Kaggle, Github is also a place where people share code so this is also a good place to learn and then apply to really learn (makes it less likely to forget),” adds Martijn.
“I advise joining an institution that can provide you with the needed certification, which will help you go beyond simply being a data developer towards earning the skills of a data scientist or a data engineer,” says Josko. “This still doesn’t come in the place of a degree to attract the attention of recruiters.”
How to get machine learning experience already during my studies?
There are different ways to get MLE experience already as a student. According to Hella, if your university gives you the option to do your thesis at a company, definitely choose to go with this option - unless you intend to go into the research field after graduation.
“I think that it is a nice opportunity to, in a sandbox, learn how it is to be a machine learning engineer,” she adds.
For Pelle and Martijn, ways to build experience during student years include Kaggle Competitions and FruitPunch AI Challenges, internships during university studies, or even enrolling in courses in computer science for example.
How can you differentiate yourself in the machine learning engineering job market?
Here are our experts advice about how to make sure that you stand out in the machine learning engineering job market:
Experience: When it comes to hardcore coding, the more your years of experience you got, the more you’ll stand out.
Communication skills: Even if you are a prodigy in building models, you need to be able to convince your client that a model is of value to them, and explain the reasoning behind your models. This requires good communication skills.
Trying new things: Distinguish yourself from others by familiarizing yourself with some of the many new trends related to machine learning operations (MLOps) where you have a variety of approaches regarding how to put a model in production. You can also stand out by going for an approach that you might have designed yourself or borrowed from other approaches.
Being critical: Act like a consultant at your project rather than a mere executioner. Challenge what you can do or what the business says you have to do.
Writing code that tells a story: Write functions and give them names that are easy to interpret, put commas to your code, and, if you’re doing a notebook, add text to your code and keep communicating what you’re doing transparently. This way you will show that you know what you’re doing.
Being focused: In an ever changing industry, it is important to admit that you cannot know everything. Read blogs to stay up to date on your niche or focus, but do not try to know everything.
Benefitting from others: After acknowledging that you cannot know everything, make use of your colleagues and their expertise to bridge any gaps in your knowledge. “Working at Xomnia, we have a lot of social activities that revolve around sharing knowledge with your colleagues and hearing about developments like these,” adds Pelle.
What is the future of machine learning engineering?
Our experts agreed that in the near future, machine learning engineering will continue to be one of the most attractive jobs in the data field, and that it will remain in high demand, especially the data engineer-related tasks in MLE.
“This particular demand for data engineering skills is partly due to the fact that when the hype around AI started, a lot of companies jumped straight to hiring data scientists to create some kind of model with the available data. Unfortunately, in many cases, the data pipelines were not set up properly, which means that the data coming into a model is not always accurate or reliable,” explains Daniela.
Here are our experts top predictions when it comes to the future of machine learning engineering:
More MLOps and automation
“In the near future, we will see more advancements in MLOPs,” says Hella. “Also, I predict that ML will be automatic and more focused on efficiency, reproducibility, and maintainability.”
DS & DE combination will change
The increase in automation will change roles in the DS & DE domain. However, there still will be a need for experts on topics that are hard to automate, leading to a shift to either data science or data engineering, but not both, according to Martijn.
MLE as a means rather than an end
"I think the future isn’t about machine learning itself. You still see companies working on fixing descriptive data platforms, which is quite archaic. I think ML will help those people to move into the space where the predictions of machine learning help you plan for the future, for instance in doing budget planning and allocations using machine learning results. That’s going to cause the real shift, which tells you that the maturity of the data industry has a long way to go.
The private sector will catch up with the market before universities do
Currently, there is a fierce competition on tech talent.
“Universities should address this competition by developing programs that are specifically tailored to forming the next generation of machine learning engineers," says Daniela. Private companies, such as AWS re/Start, are already attempting to close the gap by developing training and re-skilling programs.
"I personally have more faith in the private sector developing internal training programs to bridge the gap between what is taught in university and the skills needed to succeed in the workplace and in the future,” she adds.
How do you become a machine learning engineer at Xomnia?
If you’re interested in applying to Xomnia’s Machine Learning Development Program, or to one of our other vacancies in general, here are our experts’ advice during the application process:
Show motivation and enthusiasm
Show your passion in some way in how you would apply ML.
Show communication skills
"My favorite interview question is always 'Choose an ML model and explain it to me like I am your grandmother'", says Hella. "It doesn’t matter if it is complex or not. You need to be able to explain it in a simple way. There also needs to be a good basis in programming and statistics, but those are the obvious ones."
Be honest and be yourself
Don’t try to upsell yourself too much. Allow the interview process to determine if there is a match.
Justify your choices
Show that you know the basics of MLE by being ready to justify the steps you have taken to complete an assessment, and why you did something or not. Be very thorough.
Stick to basics
When doing your assignment, stick to the basics and what you’re familiar with, because a strong control of the basics will give a better image than messing up something advanced.
“Do not do something fancy for your assignment that you’re not sufficiently familiar with. Stick to scikit-learn, future exploration, data cleanup, understanding what’s going on,” adds Josko. “Ask yourself ís what I am doing logical?” and “does what I am doing have side effects that I didn't foresee?”