Hiring guide for ML Engineers

ML Developer Hiring Guide

ML, short for Meta Language, is a general-purpose functional programming language. It was developed in the early 1970s at the University of Edinburgh for theorem proving and symbolic computation applications. ML is known for its combination of type inference and polymorphism, garbage collection, exception handling mechanism and pattern matching. This language has been influential in the design of numerous other languages such as Haskell, Scala and F#. Today's most commonly used versions are Standard ML (SML) and OCaml.

Ask the right questions secure the right ML talent among an increasingly shrinking pool of talent.

First 20 minutes

General ML app knowledge and experience

The first 20 minutes of the interview should seek to understand the candidate's general background in ML application development, including their experience with various programming languages, databases, and their approach to designing scalable and maintainable systems.

What are the different types of machine learning?
The different types of machine learning are Supervised Learning, Unsupervised Learning, Semi-supervised Learning, and Reinforcement Learning.
What is the difference between supervised and unsupervised learning?
Supervised learning uses labeled data to predict a label given certain features. Unsupervised learning, on the other hand, uses unlabeled data and the machine learns through the discovery of certain patterns and structures in the data.
How would you handle missing or corrupted data in a dataset?
You could find missing/corrupted data in a dataset and either drop those rows or columns, or decide to replace them with another value. In Pandas, there are two very useful methods: isnull() and dropna() that will help you find columns of data with missing or corrupted data and drop those values. If you want to fill the invalid values with a placeholder value, you could use the fillna() method.
Describe the difference between a validation set and a test set.
The validation set is used to prevent overfitting of the model. It is a set of examples used to tune the parameters of a classifier. The test set, on the other hand, is used to test the performance of a trained machine learning model.
How would you handle an imbalanced dataset?
Imbalanced datasets can be handled by resampling the dataset, generating synthetic samples, or by using different evaluation metrics. Resampling can be done in two ways: undersampling your majority class, or oversampling your minority class. Synthetic samples can be generated using methods like SMOTE or ADASYN.
The hiring guide has been successfully sent to your email address.
Oops! Something went wrong while submitting the form.

What you’re looking for early on

Does the candidate have a solid understanding of machine learning algorithms?
Has the candidate demonstrated problem-solving skills?
Is the candidate proficient in programming languages relevant to machine learning, such as Python or R?
Does the candidate have experience with ML libraries and frameworks like TensorFlow, PyTorch, or Scikit-learn?

Next 20 minutes

Specific ML development questions

The next 20 minutes of the interview should focus on the candidate's expertise with specific backend frameworks, their understanding of RESTful APIs, and their experience in handling data storage and retrieval efficiently.

What are the applications of machine learning in our daily life?
Machine learning has several applications in daily life including recommendation systems like those used by Netflix and Amazon, voice assistants like Siri and Alexa, email filtering, and fraud detection among others.
What is the difference between Bagging and Boosting?
Bagging is a method of merging the same type of predictions from different models, while boosting is an iterative technique which adjusts the weight of an observation based on the last classification.
How would you explain an ROC curve to a non-technical team member?
An ROC curve is a graphical representation that shows the performance of a binary classifier system. It is created by plotting the true positive rate against the false positive rate. The area under the curve (AUC) can be used as a summary of the model skill.
What are the advantages and disadvantages of decision trees?
Some advantages of decision trees are that they require less data preprocessing, are easy to understand and interpret, and can handle both numerical and categorical data. Some disadvantages are they can easily overfit, they can become unstable due to small variations in data, and they are often relatively inaccurate.
What is the purpose of using a random forest?
Random forests are an ensemble learning method. They operate by constructing multiple decision trees during training and outputting the class that is the mode of the classes for classification or mean prediction for regression.
The hiring guide has been successfully sent to your email address.
Oops! Something went wrong while submitting the form.

The ideal back-end app developer

What you’re looking to see on the ML engineer at this point.

At this point, a skilled ML engineer should demonstrate strong problem-solving abilities, proficiency in ML programming language, and knowledge of software development methodologies. Red flags include lack of hands-on experience, inability to articulate complex concepts, or unfamiliarity with standard coding practices.

Digging deeper

Code questions

These will help you see the candidate's real-world development capabilities with ML.

What does the following Python code do?
def add_numbers(a, b):
    return a + b

print(add_numbers(2, 3))
This Python code defines a function named 'add_numbers' that takes two arguments, adds them together, and returns the result. It then calls this function with the arguments 2 and 3, and prints the result. The output will be 5.
What will be the output of the following Python code?
x = 10
y = 20
print(x == y)
This Python code compares the variables x and y using the equality operator '=='. Since x is 10 and y is 20, the comparison is false. Therefore, the output of the code will be 'False'.
What does the following Python code do?
import numpy as np
arr = np.array([1, 2, 3, 4, 5])
This Python code imports the numpy library, creates a numpy array 'arr' with five elements, and then prints a slice of this array from index 1 to 3. The output will be '[2 3 4]'.
What does the following Python code do?
import threading

def print_numbers():
    for i in range(1, 11):

def print_letters():
    for letter in 'abcdefghij':

thread1 = threading.Thread(target=print_numbers)
thread2 = threading.Thread(target=print_letters)


This Python code creates two threads using the threading module. The first thread executes the 'print_numbers' function, which prints numbers from 1 to 10. The second thread executes the 'print_letters' function, which prints the first ten letters of the alphabet. The 'start' method starts each thread, and the 'join' method ensures that the main program waits for both threads to finish before it continues.

Wrap-up questions

Final candidate for ML Developer role questions

The final few questions should evaluate the candidate's teamwork, communication, and problem-solving skills. Additionally, assess their knowledge of microservices architecture, serverless computing, and how they handle ML application deployments. Inquire about their experience in handling system failures and their approach to debugging and troubleshooting.

How would you evaluate a logistic regression model?
A logistic regression model can be evaluated using metrics such as accuracy, precision, recall, F1 score, and AUC-ROC curve. You can also use a confusion matrix to understand the performance of your model.
What is the difference between stochastic gradient descent and batch gradient descent?
Stochastic gradient descent (SGD) updates the weights using one training sample at a time, while batch gradient descent calculates the gradient using the whole dataset. SGD is faster and can be used for larger datasets, but batch gradient descent, while slower, provides a more stable and better convergence.
What is the role of the activation function in a neural network?
The activation function introduces non-linearity into the output of a neuron. This allows the neural network to learn from the error and adjust the weights, helping the model to understand complex patterns.

ML application related

Product Perfect's ML development capabilities

Beyond hiring for your ML engineering team, you may be in the market for additional help. Product Perfect provides seasoned expertise in ML projects, and can engage in multiple capacities.