← American Express Interview Insights

American Express·AI Engineer·Technical Phone Screen·Senior

Senior
May 2026

Summary

Technical screen for an AI Engineer role at Amex. One meaty conceptual question about how LLMs, RAG, and agents relate to each other. Felt like they wanted to see if you actually understood the architecture or were just name-dropping buzzwords.

Questions Asked (1)

Q1

Walk me through the differences between a base LLM, a RAG system, and an autonomous agent. How do they relate to each other, when would you use each, and how do they compose together?

System DesignTechnical Trade-offsAdaptability & Ambiguity
Author's notes

This question is deceptively wide.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

Define each system clearly, then contrast them along dimensions like control flow, knowledge freshness, and complexity. Explain how they compose hierarchically, and tie your answer to practical use cases, especially those relevant to American Express like customer service, fraud detection, and personalized recommendations.

Pro tip: Emphasize that these are not mutually exclusive but form a spectrum of increasing autonomy and complexity, and that the right choice depends on the task's requirements for accuracy, latency, cost, and maintainability.

1. Define the three systems

Briefly define base LLM, RAG, and autonomous agent, highlighting their core capabilities and limitations.

2. Compare along key dimensions

Contrast them on knowledge freshness, factual accuracy, control flow, cost, latency, and complexity.

3. Explain relationships and composition

Describe how RAG augments a base LLM, and how an agent can use both as tools, forming a layered architecture.

4. Provide use cases and selection criteria

Give examples of when to use each, considering factors like task criticality, need for real-time data, and required autonomy.

5. Tie to American Express context

Relate the concepts to potential applications at Amex, such as customer support, fraud detection, and personalized financial advice.

Key Points to Mention

  • Base LLM: pre-trained, static knowledge, no external data access, may hallucinate.
  • RAG: retrieves relevant documents to ground LLM responses, improves accuracy and freshness.
  • Autonomous agent: uses LLM for planning, can call tools (including RAG), execute actions, and adapt.
  • Composition: agents can use RAG as a tool; RAG uses a base LLM; agents may also use base LLM directly.
  • Trade-offs: base LLM is cheap/fast but less accurate; RAG adds retrieval cost/latency; agents add complexity and potential for error propagation.
  • Use cases: base LLM for creative tasks; RAG for Q&A over proprietary data; agents for multi-step tasks like transaction dispute resolution.

AI-generated suggestions, not part of the candidate's original notes. May be inaccurate — verify before relying on them.