From a Mechanical Engineer to an ML Engineer: A Career Transition

BMW, a level 5 autonomous vehicle capable of self-drive without human intervention by 2021. The BBC Project called “Talking with Machines” an interactive radio drama aims to explore the best possible ways how these devices interact with Alexa or Siri.

Machine learning is in demand!

Yes, you heard it, and why not? ML specialists, ML engineers, and AI specialists commend to be one of the best jobs in the US. Machine learning engineers applauds to earn whopping salary packages. The job role at times can pull off mind-blowing paychecks. For those with highly specialized skills, the numbers could be astronomically high. In the SF Bay area, the salary packages for machine learning engineers could range between USD 153K — USD 162K.

According to reports by “Zion Market Research” the global machine learning market predicts to rise to USD 20.83 bn by 2024, a CAGR growth rate of 44.06 between 2017–2024.

This is your cue ball.

Technology is moving pretty fast. From where the future is viewed today, it is the perfect timing to make that transition — a mechanical to a machine learning engineer.

The big-shift

The era of machines has now begun. And looking at where this technology can take us, it is a good call to making this shift. Who says mechanical engineers aren’t designed to become machine learning engineers? The transition may not be easy, but it isn’t impossible.

Being from a mechanical engineering background it is easy to comprehend what you already know and what you need to know. Without wasting much time, let us proceed.

As a mechanical engineer, you do come in with a package. However, this may not always serve the purpose of getting into another career. As a result, aspiring ML specialists or ML engineers will have to divert their focus toward machine learning tools — the latest and the most recent.

Before moving ahead, let us brush up the skills you already possess as a mechanical engineer: -

• Certain tools are limited, for instance, hardware design and machine control. Well, these are tools you do not need to frequently update.
• Technologies such as PID control and Kalman Filter are still being used and might probably stay for a longer time in the industry. Hence, you do not need to update these skills either.
• The job role can be monotonous since most of the problems are well-defined with solutions that are already tested. The only tedious task of the job role is to find the right tool and use it for the right problem.

However, this is not the case with a machine learning engineer.

As a mechanical engineer looking to hop into a machine learning career, you may want to match your skills aligned with the industry’s needs.

👉Statistics and probability

Statistics and probability are a must, especially if planning to get into a machine learning career. A solid foundation in Bayes rule, conditional probability, and likelihood is mandatory. In addition to these, the techniques such as Bayes Nets, Hidden Markov Models, and Markov Decision Processes are said to be the hearts of machine learning algorithms. As ML specialists or an engineer, it is important to understand the fundamentals of statistics and probability.

👉Programming skills and computer science fundamentals

Most of the mechanical graduates do not possess strong programming skills. Hence, the candidate needs to have a strong background in programming and understand the fundamentals of computer science.

o Data structures — arrays, stacks, queues, trees, graphs, and multi-dimensional arrays

o Computability and complexity — NP-complete problems, approximate algorithms, big-O notation, and P vs. NP

o Computer architecture — cache, memory, deadlocks, distributed processing, and bandwidth

o Algorithms — sorting, optimization, searching, and dynamic programming

👉Applying machine learning algorithms and libraries

Implementation of machine learning algorithms can be easily done using libraries or APIs like Spark, MLlib, TensorFlow, H2O, and Theano. However, applying these models can be a little tricky. For this, the individual needs to first choose a suitable model that best fits the data that can further learn from it. The neural net, decision trees, nearest neighbor, ensemble of multiple models, and support vector machine are the models you need to choose to cite the best suitable one.

While approaching, you need to be careful about the advantages and disadvantages it may incur.

Kaggle is a good platform to start getting used to real-world problems and even solve them.

👉Data modeling and evaluation

This process helps to estimate the underlying structure of datasets and aims at finding valuable patterns (clusters, eigenvectors, and correlations). The major key role of this process is to also use a valuable strategy like sum-of-squared-errors for regression, log-loss for classification.

👉System design

By the end of the process, the candidate’s core goal is to produce a positive output. Most often it is seen that small components can at times fit into larger ecosystems or perhaps services and products. Hence, the AI/ML engineer needs to draw it the smaller pieces together, communicate, and build an appropriate interface component for others to depend upon.

The duration for the transition

The time duration depends from person to person.

However, for any kind of investment, it is important to consider the return of interest. For certain people, the duration may come naturally to them — they may be able to complete their transition within months or perhaps years.

If you’re a working professional, it may be a bit tedious since you need to maintain a balance between work and upgrading your skills.

It also depends on the hours you dedicate to learning these skills.

Making a transition is not an easy task unless you have a mentor. The good news is, you will find several AI/machine learning engineer certification programs available online. Your only task is to look for a credible and relevant online platform.

Crucially, though, we all must be prepared for technology to take over!