Build a Small Agent or RAG Tool with the Mistral API
Company: Mistral AI
Role: Machine Learning Engineer
Category: ML System Design
Difficulty: medium
Interview Round: Technical Screen
In a live coding interview, you receive an API token for an LLM backend and are asked to build a small agent or retrieval-augmented generation tool. Describe the design and implementation plan for a reliable minimal system that calls the model API, retrieves context when needed, and returns grounded answers.
### Constraints & Assumptions
- The exact API surface is not specified; design against a generic chat-completions style API.
- The tool should be small enough to build live.
- Secrets must not be logged.
- AI coding assistants are allowed in the interview format.
### Clarifying Questions to Ask
- What data source should retrieval use?
- Should the answer include citations or only synthesized text?
- Are tool calls required, or is a single retrieval step enough?
- What latency and cost limits matter?
- How should failures be surfaced to the user?
### What a Strong Answer Covers
```premium-lock What a Strong Answer Covers
```
### Follow-up Questions
- How would you evaluate answer quality?
- How would you add conversation memory?
- How would you stream responses?
- How would you switch providers without rewriting the product logic?
Quick Answer: This Mistral AI ML system design question asks candidates to design a small agent or RAG tool around the Mistral API. It prepares candidates to discuss retrieval quality, orchestration, latency, evaluation, and production boundaries for LLM-powered applications.