From Physics to ML: What the Transition Actually Looks Like
If you studied physics, chemistry, biology, or engineering, you've probably wondered whether AI/ML is a realistic career path. The short answer: yes, and the transition is shorter than you think.
Month 1-3: Foundations (Ignition)
The first phase is about building your ML foundation on top of the STEM base you already have. You're not starting from zero — you're translating skills you already possess.
What you already know that transfers directly:
- Linear algebra → neural network operations
- Calculus → gradient descent and backpropagation
- Statistics → model evaluation and uncertainty
- Scientific method → experiment design and hypothesis testing
- Data analysis → feature engineering and EDA
What's genuinely new:
- ML frameworks (PyTorch, scikit-learn)
- Model architectures (CNNs, transformers, etc.)
- Software engineering practices (Git, testing, deployment)
- The ML development cycle (train, evaluate, iterate)
Month 3-6: Specialisation (Orbit)
This is where your domain expertise becomes your superpower. Instead of learning generic ML, you apply it to problems you actually understand.
A physicist might work on time series prediction for experimental data. A chemist might build molecular property predictors. A biologist might develop protein classification models.
Month 6-12: Job-Readiness (Escape Velocity)
The final phase is about building a portfolio that demonstrates real capability, not just course completion. This means:
- A capstone project that solves a real problem
- Published code on GitHub
- Technical writing that shows you can communicate
- Interview preparation with someone who's been through it
The Honest Truth
The transition isn't easy. There will be weeks where you feel lost. Concepts that seem simple on paper will be confusing in code. Your first model will be terrible.
But you have an advantage that bootcamp graduates don't: you know how to learn hard things. You've already done it for 3-4 years. ML is just the next hard thing.
