System Design Bank
🚀 AI Engineer Journey — Plan & Trackers · Hope AI — ML & DS Course
Starts: Week 5 · 1 question per session · Draw + talk approach
💡 Framework: How to Answer Any System Design Question
- Clarify requirements (ask 3 questions first — always)
- Estimate scale (back-of-envelope math)
- High-level architecture (draw the boxes)
- Deep dive on 2 key components
- Discuss tradeoffs and alternatives
- Talk through failure modes
🤖 AI-Specific System Design Questions
| Question | Week | Status | Notes |
|---|
| Design a simple conversation storage system | 5 | ⭕ | Memory types, scale |
| Design monitoring for a RAG system | 6 | ⭕ | LangSmith, latency SLAs |
| Database choice for a RAG system | 7 | ⭕ | SQL vs NoSQL vs Vector DB |
| Architecture of deployed RAG bot as API | 8 | ⭕ | FastAPI + Docker + HuggingFace |
| Design a multi-agent system | 10 | ⭕ | Supervisor-worker pattern |
| MCP at scale | 11 | ⭕ | Anthropic's design decisions |
| Production RAG system | 12 | ⭕ | Hybrid search, reranking |
| Fine-tune vs RAG vs few-shot decision | 13 | ⭕ | Decision framework |
| Production LLM serving system | 14 | ⭕ | Latency, cost, scale |
| Vector database at scale | 15 | ⭕ | Pinecone vs Weaviate vs Chroma |
| Safe AI application with guardrails | 16 | ⭕ | Constitutional AI |
| Design ChatGPT/Claude at 10M users | 18 | ⭕ | Full scale AI system |
| RAG for 100M-document enterprise | 18 | ⭕ | Chunking, indexing, retrieval |
| Multi-agent pipeline with human-in-loop | 18 | ⭕ | Agent orchestration |
| Real-time AI coding assistant | 18 | ⭕ | GitHub Copilot style |
🏗️ Traditional System Design Questions
| Question | Week | Status | Key Concepts |
|---|
| Design URL Shortener | 17 | ⭕ | Hashing, caching, CDN |
| Design Rate Limiter | 17 | ⭕ | Token bucket, sliding window |
| Design Notification System | 17 | ⭕ | Queues, fanout, push vs pull |
📚 Key Concepts Reference
| Concept | When to Use |
|---|
| CAP Theorem | Any distributed system tradeoff |
| SQL vs NoSQL | Data storage decisions |
| Redis | Caching, session storage, pub/sub |
| Message Queues | Async processing, decoupling |
| CDN | Static assets, global latency |
| Load Balancers | Horizontal scaling |
| Horizontal vs Vertical Scaling | Traffic growth questions |
| Circuit Breakers | LLM failure handling |