SciPy interviews test stats literacy, optimization basics, linear algebra, and when to use sparse/FFT—common after Python, NumPy, Pandas, and Data Science.
High-frequency topics
- SciPy vs NumPy division of labor
- p-value interpretation and test assumptions
- curve_fit / minimize workflow and local minima
- solve vs inverse; conditioning and residuals
- SVD intuition; sparse when density is low
- FFT Nyquist; filtfilt zero-phase
Whiteboard pattern
- State the scientific question (test? fit? solve? integrate?)
- Describe data as ndarray shape and dtype
- Name submodule and function
- Mention assumptions and validation (plots, residuals)
Sample verbal answer
"I'd export clean groups from Pandas to NumPy, run Welch's t-test with scipy.stats, report statistic and p-value, and check sample sizes and normality assumptions before claiming significance."
Important interview questions and answers
- Q: p-value mistake?
A: Saying p=0.03 means 3% chance null is true—wrong; it's about data extremity under null. - Q: Why not invert matrix?
A: Use linalg.solve—faster and numerically stabler than inv(A) @ b.
Self-check
- Name four common SciPy interview topics.
- Explain SciPy vs NumPy in one sentence.
- How do you validate a curve_fit result?
Tip: Practice explaining p-values, local minima, and solve-vs-inverse without jargon overload.
Interview prep
- Top topics?
Stats tests, optimize pitfalls, solve vs inv, SVD/sparse, FFT/filter intuition.
- p-value mistake?
Never say p is probability null is true.
- Local minima?
Optimizers find local minima—multiple starts and plots.