From a Physics Degree to Data Science: A Weird, Wonderful Path
I studied physics. I now work with data and machine learning. When people hear this, they usually ask one of two things: "Was it a waste?" or "How did you switch?" The answers are "absolutely not" and "slowly, then all at once."
What physics gives you
- Mathematical comfort. Linear algebra, calculus, differential equations. When a machine learning paper says "gradient," you don't flinch.
- Modelling instinct. Physics is the art of simplifying reality into a model that's wrong in useful ways. That's also a decent definition of machine learning.
- Respect for uncertainty. Error bars, measurement noise and systematic errors are taught early. Many data problems are really measurement problems.
- Estimation. "Roughly how big is this?" before computing anything. Fermi estimates catch bugs and nonsense quickly.
- Problem decomposition. Physics problems trained you to break the impossible into steps.
What it doesn't give you (usually)
- Industry-grade programming. Lab scripts that work once are different from code that a team maintains for years.
- SQL and databases. Almost every data job uses them; almost no physics curriculum teaches them.
- Business context. Revenue, churn, customer behaviour, and the art of explaining results to people who don't care about methods.
- Software engineering habits. Version control, testing, code review, deployment.
How to make the jump
1. Learn Python and SQL properly. Not "enough to plot a graph," but enough to clean messy data, write functions, use pandas fluently and write joins in your sleep.
2. Translate your experience. A physics lab project is a data project: you collected data, cleaned it, modelled it, quantified uncertainty and communicated results. Describe it that way on your resume.
3. Build two or three real projects with real, messy data, ideally in a domain you'd like to work in. One good project explained clearly beats ten tutorial notebooks.
4. Learn statistics the practical way. Hypothesis testing, regression, experiment design and the common traps. Physicists often know the maths but haven't practised the applied statistics businesses use daily.
5. Pick up engineering basics. Git, writing tests, packaging code, a little cloud. These separate "can analyse data" from "can be trusted with production."
6. Consider a structured programme if you like structure. I found short, focused professional programmes useful for filling specific gaps quickly, but they're a supplement to projects, not a replacement.
The underrated advantage
Physicists ask "does this make sense?" constantly. When a model predicts that a customer will buy 4,000 kettles, someone with a physics habit of sanity-checking magnitudes catches it immediately. That instinct, more than any equation, is what made the switch feel natural.
Advice for anyone switching from any field
Your previous field isn't wasted time; it's domain knowledge. A former biologist in health analytics, a former economist in pricing, a former teacher in education technology, they all bring context that pure data people don't have. Don't hide where you came from. Make it your edge.