Is Machine Learning Still Worth Learning in the Age of LLMs?
The essential data science concepts you need to master
LLMs are dominating the headlines, sparking questions about the future in every field, including data science. Several aspiring data scientists have asked me:
“Is machine learning (ML) still worth learning?”
My answer is a resounding yes.
While LLMs excel at text generation and conversational AI, traditional ML remains the dominant tool for predictive problems using structured, tabular data. In fact, ML is the backbone of countless real-world applications where precise predictions are crucial for the business. Let’s look at some examples:
💰 Risk Models in Lending: FinTechs and banks rely on ML models to accurately assess creditworthiness based on income, debt, and credit history.
💳 Fraud Model: FinTechs and e-commerce companies use ML to identify fraudulent transactions in real-time, protecting both businesses and consumers.
📈 Customer Lifetime Value (LTV) Predictions: SaaS companies leverage ML to predict LTV, enabling them to optimize marketing spend and personalize product experiences.
💹 Dynamic Pricing: Ride-hailing companies use ML to dynamically adjust pricing based on demand and driver availability.
For the past six years, I’ve worked in FinTech (lending and fraud). While we’ve talked about how LLMs can optimize ML model development processes, there’s been no discussion of replacing ML altogether.
LLMs are powerful tools to augment, not replace, core ML skills.
Building Foundational Technical Skills
When I started my data science journey six years ago, the landscape looked very different. I pieced together my foundation and technical skills through a mix of online resources - you can read more about that adventure here).
I dove into SQL and Python, studied foundational machine learning theory, and broadened my understanding of computer science principles through platforms like Coursera, edX, and DataCamp. While newer courses are available now, the core skills remain surprisingly consistent. Here’s what you need to master:
💾 SQL: Essential for pulling and cleaning data from databases – it's often the most time-consuming part of any project!
🤖 Python: The main coding language for data science. You’ll use it for everything from wrangling data with Pandas to building models.
🧠 Algorithms & Data Structures: Understanding how these work under the hood empowers you to choose the right tools and make informed architectural decisions.
📊 Data Visualization: A picture paints a thousand words! I spent countless hours perfecting visualizations with Python libraries like Seaborn and Matplotlib before LLMs came along…
📈 Statistical Foundations: A solid understanding of statistics helps you interpret results accurately and build robust models.
Learning to Solve Problems with Data Science
There are endless tutorials out there and it's so easy to get lost in them! If I could go back and give my past self one piece of advice, it would be this:
Don’t get stuck in tutorial hell.
I fell into that trap myself, spending countless hours on weeknights and weekends before seeing real progress. The turning point came when I started building - creating dashboards, designing data pipelines, and developing models. That's where the learning really accelerated.
If I were starting today, I’d still focus on building a strong foundation, but I’d incorporate LLMs as a coding companion for:
Debugging code,
Explaining errors, and
Suggesting improvements.
While LLMs are powerful tools, I always critically evaluate their outputs and verify explanations with reliable resources.
Wrapping Up: Your Data Science Journey Starts Now
Learning data science is a marathon, not a sprint. It's about building a strong foundation, embracing hands-on projects (even when they feel overwhelming), and continuously learning from your mistakes.
The rise of LLMs has changed the landscape, and you can leverage them wisely to accelerate your progress.
I’ve found immense satisfaction in using data science skills to uncover hidden patterns and solve real-world challenges at the companies I've worked at – from identifying key customer segments to predicting credit risk to detecting fraud. It’s incredibly rewarding to see your work directly impact a business.
The possibilities are endless, and the journey is yours to define. Let's go build something amazing! ✨
What problem will you tackle this weekend?
Keep building!
Claudia 👩🏻💻