← All questions
AI productstretch
How would you decide whether to use an off-the-shelf model API or train your own?
How to approach it
Compare on quality for your task, cost at your scale, speed, data privacy and how much the model is a source of advantage.
How a strong answer opens
“I'd compare the options on quality for our task, cost at our scale, speed, privacy, and whether the model itself is a source of advantage. In most cases I'd start with an existing model.”
Outline of a strong answer
Try answering out loud for a few minutes before you open this.
- Clarify the task and quality bar
- Test hosted models with good instructions and retrieval first; consider fine-tuning or an open-weight model only if that falls short
- Estimate cost at expected volume for each option
- Consider latency, privacy and control
- Consider team skills and maintenance
- Recommend starting with an API and revisiting at scale
Where this question usually goes wrong
Jumping to training your own model for control, without testing what existing models, retrieval or fine-tuning can already do.
Learn the method
Cost and latency trade-offs in Module 7, AI product interviews.
The full kit has worked answers in full, start to finish, for every round type. $129 once.
More ai product questions
- Your AI feature costs about four cents per request and users love it, but it would lose money at scale. What are your options?
- How would you use AI to improve Google Photos?
- How would you set up a human review process for an AI system that flags fraudulent transactions?
- What makes a good eval set, and how would you build one for an AI travel-planning assistant?
- What does an AI product manager do differently from other PMs?