Amazon Chatbot Experience — Interview Pack

{: .fs-9 }

Senior ML Platform / Applied AI Engineer interview preparation, built around a single end-to-end case study (MangaAssist) plus 30+ topic deep-dives that cover the full system-design and operations surface area. {: .fs-5 .fw-300 }

Open MangaAssist case study{: .btn .btn-primary .fs-5 .mb-4 .mb-md-0 .mr-2 } See the running web app{: .btn .fs-5 .mb-4 .mb-md-0 }

How to use this pack

Topic map

The same content, sliced two ways.

By role expectation

Tier 1 — Core technical depth (read these first)

• MangaAssist Architecture Deep-Dive — the case study HLD + LLD
• LLMOps — operations surface for LLM-backed services
• Evaluation Systems for GenAI — golden sets, LLM-as-judge, drift detection
• RAG / MCP Integration — retrieval, chunking, tool-use
• Fine-Tuning Foundational Models — when, how, alternatives
• Model Inference — serving, batching, quantisation
• Monitoring GenAI Systems — telemetry, drift, alerting
• AI Safety, Security, Governance — OWASP GenAI, guardrails

Tier 2 — Infra / engineering

• ECS / Fargate / Lambda — serverless and container compute
• DynamoDB — single-table design, access patterns
• Database Tradeoffs — relational vs NoSQL vs vector stores
• Docker Interview Pack — images, layers, Compose
• API Design and Testing — REST, gRPC, contract tests
• CI/CD Pipeline User Stories — pipeline patterns
• MLflow — experiment tracking and model registry

Tier 3 — Applied + operational nuance

• Cost Optimization (Offline Testing)
• Cost Optimization (User Stories)
• Performance Optimization User Stories
• Operational Efficiency Optimization
• Optimization Tradeoffs User Stories
• Offline Testing Quality Strategies
• Ground Truth Evolution
• Prompt Engineering
• Troubleshoot GenAI Applications
• Debugging

Tier 4 — Foundations

• Modeling Math Foundations
• Statistical Inference
• Statistics Foundations

Tier 5 — Domain stories + interview format

• HLD Questions
• LLD Questions
• Real-World Interview Questions
• Domain 1 — FM Integration & Data Compliance
• Implementation / Integration — Domain 2
• Applied ML Engineer User Stories
• ML Engineer User Stories
• Tech Stack
• AI-First Production Engineering
• Security / Privacy Guardrails
• Leadership Narrative
• POC to Production War Story
• Challenges
• MangaAssist Interview Pack

By interview competency

The MangaAssist case study

The numbered files at the repo root walk through the entire case end-to-end, the way a real ML platform interview unfolds:

Sibling code in this repo (not part of the docs site)

The interview-prep notes live alongside three working prototypes:

• mangaassist_web/ — Next.js 14 + FastAPI app that implements the case study's customer + admin surfaces. Source of truth for the production-shaped story.
• streamlit_app/ — Streamlit prototype with 21 pages covering every experiment (RAG bake-off, model arena, prompt studio, voice console, guardrails lab, etc.).
• mangaassist_3d/ — WebGL atlas exploration.

These folders are excluded from this docs site by config; reference them on GitHub when an interviewer asks "show me the code that does this."

{: .tip }

Every folder page on this site has the same shape: 1-paragraph summary, "interview talking points" (the questions this folder helps you answer), and an auto-generated table of contents over every .md file in the folder. Use the search box (top of every page) when you have a specific term in mind.