MangaAssist Interview Pack - Basic With Hints

Level: Basic
How to use: Answer out loud first, then use the hints to check whether you covered the project-specific points.

Memory Map

graph TD
    A[Business Goal] --> B[Customer Needs]
    B --> C[Use Cases]
    C --> D[Core Architecture]
    D --> E[Safe Response]

Interview Questions With Hints

Hiring Manager

1. Tell me what problem MangaAssist is solving for the JP Manga store.

Hint: Focus on discovery overload, information gaps, fragmented support, and lost conversions.

1. Who are the main users of this chatbot, and how do their needs differ?

Hint: Cover new visitors, existing Amazon customers, manga enthusiasts, comparison shoppers, support seekers, and international buyers.

1. Why is this system more than just a FAQ bot?

Hint: Mention live catalog, recommendation, shipping/order flows, personalization, and escalation.

Product Manager

1. Which user journeys would you call the most important for the MVP, and why?

Hint: Anchor on discovery, recommendation, FAQ, and escalation because those are both differentiated and feasible.

1. What is the north-star experience this chatbot is trying to create for the customer?

Hint: Useful answer in under 3 seconds without leaving the store page.

Senior Engineer

1. Walk me through the high-level architecture layers of MangaAssist from the store page to the final response.

Hint: Client, edge/auth, orchestration, intelligence, data, safety/output, observability, fallback.

1. Why does the system have an orchestrator instead of letting the frontend call every backend directly?

Hint: Central coordination, session context, intent routing, aggregation, guardrails, and decoupling.

1. What kinds of requests should go to direct APIs, and what kinds should go through RAG or the LLM path?

Hint: Use APIs for live structured data like order status or prices; use RAG for policy/FAQ; use richer LLM paths for ambiguous or recommendation-heavy flows.

Support Lead

1. When should the chatbot answer on its own, and when should it hand off to a human agent?

Hint: Hand off on explicit request, billing/payment disputes, repeated failure, sensitive topics, or very negative sentiment.

Data Analyst

1. What are the first three success metrics you would watch after launch?

Hint: Pick one business metric, one UX metric, and one operational or AI-quality metric.

Quick Recall Diagram

sequenceDiagram
    participant User
    participant Widget as Chat Widget
    participant Gateway as API Gateway
    participant Orch as Orchestrator
    participant Router as Intent Classifier
    participant Services as APIs/RAG/Reco
    participant Safety as Guardrails

    User->>Widget: Ask question
    Widget->>Gateway: Send message
    Gateway->>Orch: Auth + route
    Orch->>Router: Classify intent
    Router-->>Orch: Intent type
    Orch->>Services: Fetch data or context
    Services-->>Orch: Results
    Orch->>Safety: Validate answer
    Safety-->>Widget: Safe response