What is dependency pinning and why does ML need it?
PICTURE THIS: HOW TO EXPLAIN IT
Simple meaning
Pinning locks package versions in requirements.txt or a lockfile so numpy or CUDA updates do not silently change results.
WHY — Reproducibility instead of guessing?
Why interviewers care about Reproducibility:
who only read docs from people who shipped.
and tied to MLOps work.
Name the idea, why it exists, then one short example.
End with when you use it and one common pitfall.
STEPS — What happens step by step?
Before you speak the answer, walk the interviewer through these steps:
- 1Pinning locks package versions
in requirements.txt or a lockfile so numpy or CUDA updates do not silently change results.
- 2ML is sensitive to
numerical libraries.
- 3Unpinned environments are a
top cause of unreproducible experiments.
- 4Give an example
One tiny concrete case you can say aloud.
- 5Common mistake
What juniors usually get wrong.
- 6Close
When you pick this over the alternative.
EXAMPLE — See it in action
Here's a short line you can speak, broken into clear beats:
Note: Adapt this scaffold to your own project — keep it under 60–90 seconds.
Key takeaway
Pinning locks package versions in requirements.txt or a lockfile so numpy or CUDA updates do not silently change results. ML is sensitive to numerical libraries.