CF
Machine Learning thumbnail

Machine Learning

21 courses 4 categories

Part of Learn Data & AI

Machine learning as a topic covers the classical and deep-learning fundamentals that underpin every modern AI system — supervised learning, neural networks, optimization, and the math beneath them. Unlike the broader AI hub or the LLM-engineering specialty, this topic zooms into the model-building side: training networks from scratch, working with tabular and image data, understanding why models converge, and shipping non-LLM ML in production. No conversational chatbots front-and-center here.

The 2026 toolchain is the same one researchers actually use. PyTorch dominates day-to-day work; JAX holds the high-end research market and TPU workloads; scikit-learn remains the right answer for tabular problems before reaching for a neural net. NumPy, pandas, and polars handle data prep; Weights & Biases and MLflow track experiments; ONNX and TensorRT cover deployment when latency matters. Computer vision pipelines still lean on OpenCV alongside modern vision transformers.

What you'll find under this topic

  • Supervised learning: regression, classification, decision trees, gradient boosting (XGBoost, LightGBM)
  • Deep learning: feed-forward, convolutional, recurrent, transformer architectures from scratch
  • PyTorch deep dives: autograd, custom datasets, distributed training, mixed precision
  • Math foundations: linear algebra, probability, calculus, optimization theory for ML
  • Computer vision: OpenCV pipelines, object detection, segmentation, image generation
  • Model evaluation: cross-validation, calibration, fairness audits, drift detection
  • Deployment: ONNX export, TorchServe, edge inference, model compression

Machine learning roles still hire across the whole industry — recommendation systems at Spotify, Netflix, and Uber; fraud detection at fintech; medical imaging at health-tech companies; perception stacks at autonomous-vehicle teams. The skill set is durable because the underlying math has not changed and the production problems (data drift, label noise, evaluation rigor) keep recurring.

Top 10 picks for 2026

Categories (4)

Deep learning thumbnail
Deep learning
Deep learning is a transformative subset of machine learning centered around neural networks with multiple layers…
Machine learning thumbnail
Machine learning
Machine learning is a subset of artificial intelligence that enables computers to learn from data and make decisions…
Math & Statistics thumbnail
Math & Statistics
Math & Statistics underpin many aspects of software engineering and data science, providing the foundational tools…
OpenCV thumbnail
OpenCV
OpenCV is a versatile computer vision library initially released in 2000 and maintained by the OpenCV community. It…

Courses (21)

Showing 1 – 21 of 21 courses

Related topics

Frequently asked questions

Do I need a degree to work in machine learning?
For pure research roles, usually yes — most ML researcher hires hold a PhD or a strong publication record. For applied ML engineering, no. Solid Python, statistics, the standard ML toolchain (scikit-learn, PyTorch), and a portfolio of real projects open most applied roles. The bachelor's-versus-PhD gap matters at frontier labs, much less at product companies.
Machine learning vs LLM engineering — which to focus on?
LLM engineering for fast time-to-impact at most product companies — the skill set leans more on software engineering than math. Classical ML for forecasting, recommendation systems, fraud, ranking, scientific computing, and any tabular-data problem where deep learning is overkill. Strong applied ML engineers know both and pick the right tool per problem.
What math do I really need for ML?
Linear algebra (vectors, matrices, eigenvalues conceptually), probability and statistics, basic calculus (gradients, partial derivatives). You don't need to derive everything from scratch; you do need to read papers and debug models without being lost. Most applied engineers refresh these progressively, picking up depth as their model work demands it.
PyTorch vs TensorFlow vs JAX?
PyTorch dominates research and an increasing share of production. TensorFlow is still common in Google-adjacent ecosystems and some enterprise. JAX is gaining ground for high-performance research and Google internal work. For a new learner the only sensible pick is PyTorch — by a wide margin the largest community and the easiest path to current research code.
How long to become an applied ML engineer?
12–24 months from a strong software-engineering baseline, longer from scratch. Plan on solid Python, a year of classical ML projects (regression, classification, clustering, time series), a deep-learning specialization, and one real production deployment with monitoring and drift detection. Kaggle competitions help build intuition but don't substitute for shipping models that someone depends on.