What You Need to Know Before Switching to a Data Science Career in 2024

Author:Murphy | View: 20630 | Time: 2025-03-22 20:13:42

This article is based 100% on my experience, my research, and what I'm observing in the European market. Things might look slightly different in the rest of the world, but we're all facing the same waves of change in the data science market.

For those who don't know me, my journey started ten years ago: I graduated with a degree in applied mathematics from an engineering schools. At that time, data science wasn't a trendy term. I kicked off my career at EDF (Électricité de France), a leader in the energy sector, building models to understand energy consumption and production patterns using time series and probabilistic models.

Back then, having a solid foundation in applied mathematics was like holding a golden ticket into the world of data science. Data science jobs were rare, and I often found myself explaining to people what it even meant to be a data scientist.

Over the last decade, the field has transformed completely. Some juniors I mentor feel that the market has become oversaturated. So, if you're considering switching to data science in 2024, I'd say this: the game has changed – but it's still very much playable if you have a clear roadmap.

By the end of this article, I'll give you the roadmap I would follow if I were starting my journey today.

Let's start by the market evolution

Let me take you back to 2015: Data science was starting to make waves. Universities were rebranding old programs with shiny new names, adding a couple of courses related to data to keep up with the trend. Bootcamps started popping up like mushrooms. In France, the government helped fund many of them to meet the growing demand for data experts.

Then came the boom years of 2018. Data science was suddenly everywhere. The world was drowning in data, and companies were scrambling to find people who could make sense of it. The number of degrees in data science exploded, and the market was hungry for talent.

And then came 2022 and 2023 – the tech world was rocked by layoffs. The atmosphere shifted. But despite the layoffs, data science remained resilient compared to other IT jobs. So, while layoffs affected the tech industry, data scientists weren't hit nearly as hard. But that doesn't mean things haven't changed. You can find all the detailed numbers about the layoffs here.

In France, we can clearly see that the demand for data scientists has changed. According to the GEN France 2024 Report (a government initiative to track job trends), the most in-demand roles today are in cloud, network, and security – not data science. You can see more details in this report: Job postings in the Security, Cloud, and Network categories accounted for more than a third of all postings. Almost a quarter of the postings were for Careers in Development, Testing, and Operations.

With all the hype around genAI and companies shouting about the future of AI, it can feel like a contradictory time. Everyone wants to pivot into AI , and yet job ads are all about cloud, dev, and operations. So, what should you do if you're looking to enter data science in 2024?

What's Behind the Numbers?

From what I'm observing, the demand for data scientists hasn't disappeared – it's just wildly evolved. Employers today expect you to know more than just the models you build. You need to know how to integrate those models into production systems, monitor them, and ensure they're bringing value to the business. If you think you can spend all day just tweaking models and tuning hyperparameters, you're in for a rude awakening.

So, what's really going on here?

It's simple. AI tools are automating a lot of the tasks that data scientists used to spend hours doing. Employers are now looking for people who bring more to the table – people who can manage machine learning in production, understand cloud architecture, and connect the dots between Data Science and business impact. Just knowing Python or machine learning isn't enough anymore.

My Advice to Future Data Scientists

1. Master the Basics

Here's the harsh reality: If terms like linear algebra, probability, or data structures don't mean much to you, stop right there. Data science isn't just about importing models from scikit-learn and watching the magic happen. You need to understand why and how those models work. It's not about knowing a lot, but really mastering what matters.

Take PCA (Principal Component Analysis), for example. You can't just run a PCA without understanding eigenvalues and eigenvectors – it's essential to know the underlying math. You also can't debug your Python code if you don't understand the data structures you're working with.

Start by nailing the fundamentals: statistics, probability, linear algebra, data structures, and mastering one programming language (probably Python). If you're still a student, take the time to master these basics during your degree. If you're going to a bootcamp, spend three to six months beforehand building a foundation in these skills. It'll save you from a world of pain later on.

I highly recommend starting with these courses:

Mathematics for Machine Learning Specialization (Imperial College): This course is designed specifically for people who want to apply math in Machine Learning. It's not about heavy theory, but about practical math that helps you in the real world. Trust me, the instructors break it down so anyone can understand.
Khan Academy's Statistics and Probability: This is your go-to for mastering the basics of probability and statistics. No fluff, just the core concepts you'll use every day in data science.
Harvard's Introduction to Calculus: Don't let the name intimidate you – this course starts simple, and by the end, you'll have a solid grip on the derivatives and optimisation techniques that you need to understand your machine learning algorithms.
Python for Everybody Specialization: With this course you will learn to Program and Analyze Data with Python.

Each of these resources is free (or accessible as a free auditor), and they will provide you with the essential skills to build on. By investing just a few months here, you'll save yourself countless headaches down the road.

2. Leverage Your Existing Experience by building your portfolio

One of the biggest advantages you have is your existing field knowledge. If you come from biology, finance, or marketing, you're already one step ahead of those who don't understand the domain they're working in.

For example, a biologist diving into data science can bring unique insights into biological data that a generalist data scientist wouldn't have. If you've been working in finance, you already know the important metrics and data that are valuable in your field. The trick is to combine your domain expertise with data science techniques.

Start by working on personal projects that merge the two. Build a portfolio that showcases your ability to solve real-world problems in your specific field using machine learning.

Need inspiration? Check out my article: 8 Projects to Skyrocket Your Data Science Portfolio.

The best way to build your Portfolio:

Data Exploration: Start by discovering, analyzing, and cleaning data using essential Python libraries like pandas and NumPy. This step will make you comfortable working with data exploration tools.
Data Visualization: Create insightful visualizations using libraries like Matplotlib and Seaborn, then take it further by building a dashboard using Streamlit or Dash. This will show off your ability to communicate data effectively.
Build a Machine Learning Pipeline: Develop a pipeline from start to finish – choosing the right model, training, and evaluating it. This is where you bring machine learning to life.
Interpret Results: Go beyond the numbers. Explain the impact of your findings.
Understand the algorithms you are using with the Statistical learning course by Stanford: This is a life changing course for someone who want to start in data science and understanding in depth the most important algorithms. You also have the book version in python and R here.
Keep learning and refining your portfolio with advances techniques: the machine learning, and deep learning from courses from Coursera will introduce you to advanced techniques like deep learning (CNN for images, transformers for text, LSTM/RNN for time series, or even reinforcement learning).

3. Learn Software Engineering (Really – This Will Set You Apart)

Here's where many data scientists fall short: they can build models, but they don't know how to deploy them. In 2024, that's a deal-breaker. If you can't move your models from a Jupyter notebook to a real-world environment, you're missing a huge part of the puzzle.

The good news? You don't need to become a full-stack developer, but learning software engineering basics will make you unstoppable. Here's how to get started:

Create an API: Take your existing model and create an API to serve its results. This is how you prove you can move beyond notebooks and actually build something deployable in the real world. APIs are the bridge between your model and the users.
Learn Docker & Azure: If you want to take your deployment skills to the next level, get familiar with Docker and Azure. These tools help you package your models into containers and deploy them in the cloud.
I recommend you these 3 step by step tutorials: Introduction to Docker, Model Management with Azure and Docker , FastApi deployment with Azure and Docker.
Learn SQL: Once your model is deployed, you'll often need to store the results in a database. But before diving into databases, meet your new friend – SQL. SQL is a must-have skill for managing and querying data, and it's essential for anyone in data science.
Start with SQL for Data Science (UC Davis): This course will give you hands-on SQL practice, from basic queries to more complex operations. It's tailored specifically for data scientists, so you'll gain practical, relevant skills.
Practice with MySQL Workbench: The most straight forward way of interfacing with a SQL database is to use a program like MySQL Workbench which is available for mac and windows. MySQL Workbench comes with the popular Sakila database. A sample database designed to help you practice SQL queries. information on the sakila database here.
Push Your Code to GitHub: Once your API is deployed, push all your code to GitHub and set up CI/CD workflows to automate your deployment. This will not only keep your project up-to-date but also showcase your ability to handle the full development pipeline.

4. Specialization

Data science is a vast field, and while it's tempting to learn a little bit of everything, specialization is the key to your dream job. Pick a niche that aligns with your existing skills or interests – whether it's time series analysis, computer vision, NLP, or reinforcement learning. Becoming an expert in one of these areas will make you a valuable candidate in the job market.

For example, if you have experience in finance, focus on time series analysis for financial forecasting. If you're in healthcare, delve into computer vision for medical imaging. Whatever your background, tie it to a specific data science field to give yourself a competitive advantage.

There are tons of resources available to help you specialize:

DeepLearning.ai: They offer courses on cutting-edge techniques like transformers, CNNs, LSTMs, and more. These advanced courses will give you the depth needed to specialize in fields like NLP or computer vision.
Papers with code: Stay up to date with the newest advancements in machine learning. This is where you can find and experiment with the latest techniques used in real-world applications and apply them to your data.

By specializing, you'll not only become more marketable but also position yourself as an expert in a niche that employers value. And remember, the more focused your expertise, the easier it becomes to demonstrate real impact in your projects.

5. Understand the Impact of AI Tools (And Use Them Wisely)

With tools like ChatGPT and other AI-powered solutions automating more and more basic tasks, the role of a data scientist has evolved. A senior data scientist once told me, "AI isn't replacing data scientists; AI is replacing data scientists who can't bring more value than the AI itself." In other words, if all you can do is what AI can already automate, you're at risk.

For example, ChatGPT can write your Python code for you, but it can't tell you if your data doesn't meet the assumptions for a particular statistical test. That's where your expertise comes in. You need to understand the data, choose the right techniques, and know when something isn't quite right. Your critical thinking skills and domain knowledge are what differentiate you from AI.

Here's how you can use AI tools to enhance your work, not replace your critical thinking:

Use AI tools to boost productivity: ChatGPT and similar tools can help you write code, perform quick calculations, or even summarize research papers. These tools save you time on repetitive tasks, allowing you to focus on more complex problems.
Know when to step in: AI is great, but it's not perfect. Your role is to validate the AI's output, ensure data integrity, and make decisions that AI can't handle. For instance, statistical analysis requires human judgment, especially when determining which test is appropriate for your data.
Understand AI ethics: As AI becomes more integrated into the workforce, understanding the ethical implications of AI is crucial. This includes knowing how to avoid bias in models and understanding AI's potential impact on society. For that, I highly recommend Ethics of AI, a free course created by the University of Helsinki. It offers a comprehensive overview of AI ethics and its role in responsible AI development.

As I Promised: The Roadmap I Would Follow If I Started My Journey in 2024

Mastering the Basics → Mathematics for Machine Learning Specialization (Imperial College) → Khan Academy's Statistics and Probability → Python for Everybody Specialization → Harvard's Introduction to Calculus

Mastering Algorithms by Building a Portfolio → Building a project from scratch: starting from data exploration to ML models using Numpy, Pandas, Matplotlib, Seaborn, Scikit-learn. 10 Great Places to Find Free Datasets for Your Next Project → Machine Learning by Coursera → Statistical learning by Stanford

Mastering the Basics of Software Engineering → Introduction to Docker → Model Management with Mlflow, Azure and Docker → FastAPI Deployment with Azure and Docker → SQL for Data Science (UC Davis) → GitHub and CI/CD → Building a dashboard with Streamlit

Learning How to Specialize by Mastering Advanced Concepts → DeepLearning.ai → Papers with Code

Understanding AI Ethics and Human Impact → Ethics of AI

I've mentored hundreds of junior data science students and gathered their input on what truly helped them succeed in the market. if you follow these steps, it'll make your journey smoother.

Keep learning, stay positive, and you'll do great! Good luck!

Thank you for reading!