Hire Machine Learning Engineers

Machine Learning Engineers Are a Modern Business Necessity

With 91.6% of Fortune 1000 companies increasing their investments in big data and artificial intelligence, it is no surprise that the demand for machine learning (ML) engineers is also growing significantly and showing no signs of slowing. In fact, LinkedIn had ML engineers listed as 2022’s fourth-fastest-growing role in the US alone, and the job market for ML engineers is projected to increase by 31% through 2026.

So how hard can it be to find an ML engineer? Well, not very hard at all if the goal is just to find someone who can legitimately list machine learning on their resume. But if the goal is to find an ML expert who has truly mastered its nuances, power, and strategic applications, then the challenge is most certainly formidable.

You will need to understand both your business needs and how ML may be used to implement the ideal solutions. You’ll then want to create a highly effective recruiting and evaluation process specifically geared toward finding not simply a qualified ML engineer, but the right ML engineer for your specific needs. Your first move in this process, however, is to read on and learn more about each of these critical steps.

What attributes distinguish quality Machine Learning Engineers from others?

Talented ML engineers are not only theoretically knowledgeable and technically proficient but also possess a variety of soft skills that enhance their ML-specific abilities.

A quality ML engineer should possess:

• Theoretical knowledge
  • Basic and ensemble ML models
  • Regression, classification, clustering, reinforcement learning, optimization
  • ML metrics appropriate for different model types
  • Deep learning (DL) architectures for different applications
• Technical skills
  • Good coding skills: Python and/or R, SQL, and experience with other software engineering languages is a plus.
  • MLOps tools and platforms: These will depend on your existing or desired infrastructure, e.g., SageMaker, AzureML, H2O.ai, Vertex AI, DataRobot, etc.
  • Big data technologies and tools: Parquet files, big file viewing/editing tools
  • Visualization: Matplotlib, seaborn (good-to-have: sweetviz, plotly)
  • Databases (both relational and non-relational) and data warehousing solutions
• Soft skills
  • Ability to understand and solve problems with minimal guidance from the business
  • Ability to question assumptions
  • Investigative mindset and data-driven argumentation

How can you identify the ideal type of Machine Learning Engineer for you?

You now know how to identify a quality ML engineer from a general standpoint. However, ML problems can be quite varied, so you’ll need to identify your specific business needs in order to find the ideal ML engineer to address them. Start by drafting a “problem statement” to identify the issues you’re looking to solve and how ML will be a part of the solution.

Your problem statement should include, at a minimum, the following considerations:

• Identify problems
  • What business cases are you looking to improve?
    • Some business cases for ML considerations can be found here, with additional insights below.
  • Are you looking for a long or short-term engagement?
  • Do you have a well-defined requirement, or are you looking for someone to help with the business process overall using ML?
• Define stakeholders
  • Which areas of the business require the expertise of an ML engineer?
  • Who will be available to participate in the design/redesign of ML-empowered processes?
• Define technologies
  • What are your existing/desired cloud/on-premises platforms?
  • What programming languages are used in your business?
  • What databases do you have or plan to use?
• Consider MLOps
  • What level of automation does the project require?

Once you’ve addressed these questions in your problem statement, you can use the following guides to determine 1) whether your needs are best suited to a junior or senior-level ML engineer and 2) the particular candidate skill sets you should prioritize, based on your specific business cases:

Experience Level

Junior ML Engineers: These engineers will be able to make decisions in the areas of data selection/preparation, model development, and technology implementation. They’ll also be expected to take guidance from your data scientists and DevOps engineers.

Senior ML Engineers: By virtue of their more extensive backgrounds and longer histories in the space, quality senior ML engineers will likely be more advanced in the day-to-day functions noted above. However, they will also look beyond the day-to-day with a “big picture” mindset to identify the areas of your business that can be improved using ML. Senior engineers should be able to understand your business process and deliver high-quality, scalable solutions that integrate seamlessly with your existing infrastructure.

Priority Skills by Business Case

Forecasting: Look for an ML engineer who understands time series models. Sophisticated predictive models such as Prophet or Long Short-term Memory (LSTM) offer good performance but sometimes hide the complexity of the underlying data. To ensure that your data is well-explored, look for an ML engineer who understands the basics of time series, i.e., seasonality, trend, autoregressive properties, and stationarity.

Customer segmentation: Look for an ML engineer who has knowledge of clustering algorithms, techniques for defining the number of clusters, and performance of clustering models. A good understanding of business metrics, such as customer satisfaction, purchase history, and customer lifetime value, is also important.

Fraud detection: Look for an ML engineer who has experience with anomaly detection models, unsupervised learning for detection of new fraud patterns, unbalanced classification and/or clustering, and understanding outliers, as well as effective application of ML metrics to maintain model relevance.

Identity verification, video monitoring, and/or automatic video and image labeling: Look for an ML engineer who understands that many systems require the examination of video streams, e.g., to identify intruders on a property, to assist in remote identity verification, to automatically classify movies and TV shows by genre, or to detect actors. These skills rely on image classification/segmentation techniques, which are often based on state-of-the-art deep learning architectures, so understanding them is essential. However, it is also important for an ML engineer to understand the intricacies of video stream processing, data compression, storage of large unstructured data, and performance of ML models trained on images.

Sound classification and voice generation: Look for an ML engineer who understands that sound is typically processed using the Fourier transform to create time/frequency “images” that can be processed in the same manner as images. Additionally, the right candidate will also possess the image classification/segmentation skills and experience noted in the identity verification skills description above.

Text processing, chatbots, search engines, and text generation: Look for an ML engineer who has expertise in text tokenization, embedding, simple models (such as Multinomial Naive Bayes and Word2vec), and state-of-the-art models. In addition to modeling, experience with text storage and compression is also important.

How to Write a Machine Learning Engineer Job Description for Your Project

You’ve identified the experience level and skills you will require in your ML engineer. Now it’s time to find that perfect fit. Like most job postings, each job title will feature a similar/standard set of roles and responsibilities. To aid in this piece, consider referencing this ML engineer job posting template. Then, make sure to include explicit requirements as determined in your problem statement considerations to help candidates self-select before applying.

What are some key interview questions to ask Machine Learning Engineer job candidates?

As much as both interviewer and interviewee will often try to stick to a basic script, most good interviews will veer into conversational territories as each answer will often raise further unplanned, but related follow-up questions. That said, it’s important to have a list of questions that you know you’ll need answered to properly assess your candidate. Consider having these interview questions at the ready during your meeting, as well as the following:

What is the difference between deep learning and machine learning?

The main difference between deep learning (DL) and ML is that DL is a neural network (NN) architecture with multiple hidden layers. While conceptually this is not much different from a single-layer NN (an ML perceptron), the addition of hidden layers allows for the encoding of very complex relationships between features and a target variable. This, in turn, allows for efficient processing of large unstructured data, such as images, text, and audio.

What types of neural networks exist?

The field of artificial neural networks is constantly evolving and many types of NNs have been proposed, tested, and put into practice. Although different sources propose different taxonomies, the most common types of NNs in typical commercial applications include the perceptron, feed-forward NN, convolutional NN, and recurrent NN. Perceptrons are useful for creating basic models. Feed-forward NNs have applications in various fields, and their advantage is the availability of different activation functions. Convolutional NNs are widely used in image-, video-, and sound-processing applications. Recurrent NNs are used in sequence processing, most notably in NLP. They are the basis of transformer neural networks.

How many of the top 10 ML algorithms can you name? How familiar are you with each? Please give examples.

Linear regression, logistic regression, K-means, random forest algorithm, SVM algorithm, decision tree, KNN algorithm, Naive Bayes algorithm, gradient and AdaBoost algorithm, dimensionality reduction algorithms.

Why do companies hire Machine Learning Engineers?

Machine learning engineers use ML to improve data processing and insight extraction. ML engineers share many responsibilities with data scientists; however, in addition to building models, ML engineers develop pipelines and maintain models in a production setting. Their typical workflow includes the following steps:

• Data collection: optional, can be performed by data engineers or data scientists, or can be taken from a ready data set
• Exploratory data analysis (EDA)
• Base model building
• Best model selection: comparison of models, cross-validation, hyperparameter tuning
• ML pipeline creation: set up of evaluation metrics, alert methods, and integration with business apps and dashboards to support cost-effective model deployment and monitoring.
• ML pipeline testing
• ML pipeline deployment: docker, Terraform script, or similar
• Performance tracking and maintenance: automated retraining, performance alerts, and experiments tracking
• Model upgrades: usually when new types of data become available or business objectives are modified

Conclusion

Increasingly, machine learning is providing solutions to many of our everyday issues, both personal and professional. As business leaders, integrating ML into as many aspects of our work as possible is not simply sufficient; it’s necessary to get and stay ahead of our competitors.

Developing a high-level understanding of ML and a proficiency in its many current and potential business applications will provide you with the critical abilities to identify ML use cases in your own company and hire the ideal ML engineer(s) to implement the right solutions.