AI System Design Guide
A living reference for production AI systems, covering RAG, LLM engineering, agentic AI, and interview prep.
- Source repo
- ombharatiya/ai-system-design-guide
- Stars
- ★ 3.6k
- Last updated
- 10d ago
- License
- MIT
- FA score
- 27/100 · Major gaps
At a glance
- How it runs
- Works with
- Universal · cross-platform
- Cost
- Free, no paid service needed
- Setup effort
- Low · running in minutes
- Typical use
- Engineers preparing for AI system design interviews can use the question bank and answer frameworks.
- Not a fit if
- Readers who need runnable code or a hands-on lab
- Absolute beginners without ML fundamentals
- Readers wanting content that stays stable over time
- Source review
- 27/100 · Major gaps 8 safety controls not found
What does this agent do, and when should you use it?
This is a continuously updated, open-source guide for engineers building production AI systems and preparing for senior-level interviews. It covers RAG architectures, LLM engineering, agentic AI, MCP and A2A protocols, model selection, evaluation, and real-world case studies. The repository includes 19 topic chapters, a 122-question interview bank, 15 case studies with diagrams, and two deep-dive evaluation guides. It tracks July 2026 model releases and pricing, contrasting its freshness with printed books. The project is MIT-licensed, open to contributions, and also available online at aidaddy.tech.
The repository is a curated collection of Markdown documents organized into topical directories. It provides structured learning paths: from LLM internals to RAG construction, agent development, model selection, evaluation, and case studies. It includes a 122-question interview bank with answer frameworks, plus nine whiteboard exercises. Two deep-dive evaluation guides cover Phoenix/Langfuse and LangWatch/Langfuse, detailing LLM-as-a-judge, RAG evaluation, and statistical correction. Each case study presents a problem, key patterns, and architecture diagrams to illustrate production failure modes. The guide is a living document, updated as new models and protocols ship.
- Engineers preparing for AI system design interviews can use the question bank and answer frameworks.
- Architects designing multi-tenant RAG systems can refer to security and access control chapters.
- Developers building AI agents can learn about MCP/A2A protocols and loop engineering.
- Technical leads selecting models for production can compare pricing and capabilities from the model taxonomy.
- Teams setting up AI evaluation pipelines can follow the two comprehensive eval guides.
- Professionals transitioning from backend, QA, or PM roles can use the transition guide and course recommendations.
How do you install or deploy this agent?
No installation required. The guide is a set of Markdown files in the GitHub repository. You can clone the repo or browse it online.
How do you use this agent?
After cloning, read the chapters relevant to your goal. The README provides a quick navigation table; for example, start with '00-interview-prep/01-question-bank.md' for interviews, or '06-retrieval-systems/01-rag-fundamentals.md' for RAG. You can read offline or use the aidaddy.tech website for search and a better reading experience.
What are this agent's strengths and limitations?
- Continuously updated with July 2026 models (Claude Opus 5, GPT-5.6, etc.) and protocols (MCP 2.0, A2A).
- Comprehensive: 122 interview questions, 15 case studies with diagrams, and two deep-dive evaluation guides.
- Provides concrete tradeoffs (latency vs. cost, accuracy vs. faithfulness) rather than generic advice.
- MIT-licensed and free, with active contribution opportunities.
- Structured learning paths and visual diagrams help navigate large content.
- No executable code or interactive environment—it's a reference document, not a hands-on tool.
- Large volume (19 chapters) may overwhelm beginners.
- Highly dependent on 2026 model landscape, which will date quickly without updates.
- Assumes prior knowledge of ML fundamentals; not for absolute beginners.
How does this agent compare with similar options?
Compared to traditional printed books, this guide is always up-to-date, avoiding dated content. It doesn't directly compete with specific tools, but rather with online courses or tutorials; the README recommends courses in COURSES.md as supplements.
Key facts side by side with the most closely related agents.
| Agent | Source review | Form / cost | Stars | Updated | Language | Full support on |
|---|---|---|---|---|---|---|
| AI System Design Guide This agent | 27 · Major gaps | Web appFree | ★ 3.6k | 10d ago | — | — |
| AgentGuide - All-in-One AI Agent Development & Interview Guide | 23 · Major gaps | Library / SDKFree + model costs | ★ 11k | 1mo ago | MDX | OpenAI API · Claude API |
| AI Agents from Zero: Practical Bootcamp to Enterprise Deployment | 34 · Major gaps | CLIFree + model costs | ★ 5.4k | 5d ago | Python | OpenAI API · Claude API |
| AI Agent Engineering Textbook | 55 · Major gaps | FrameworkFree | ★ 601 | 2d ago | HTML | — |
How does FollowAgents rate this agent?
Why each dimension lost points
Evidence shows this is a static documentation repository with no executable code, permission declarations, or data processing logic. Therefore, all trust-related criteria (least privilege, user confirmation, data flow transparency, sensitive data handling, dependency security, external effects, rollback, source attribution) have no supporting evidence and are scored 0. For source attribution, the README and LICENSE clearly identify the author (Om Bharatiya), but the publisher identity is unverified and there is no detailed provenance of content sources, so no credit is given.
Self-consistency: The guide has a clear structure, consistent chapter divisions and navigation, and FAQ aligns with content, with no obvious contradictions, scored 2. Dependency availability: The repository declares no runtime dependencies, so it cannot be assessed, scored 0. Failure messages: No error handling or troubleshooting information is provided, scored 0.
Audience and scenarios: Clearly targets AI engineers, interview candidates, etc., with multiple usage paths and scenarios, scored 2. Capability boundaries: Clearly states what the guide is and is not, scored 2. Trigger precision: No trigger mechanisms or instructions are defined, scored 0. Environment fit: As documentation, it does not depend on a specific environment, but no installation or running instructions are provided, scored 1.
Information architecture: Clear directory structure and comprehensive navigation table, scored 2. Install notes: No installation steps, scored 0. Naming stability: Chapter naming is consistent, but no versioning or change log is provided, scored 1. Examples and FAQ: Rich examples and FAQ provided, scored 2. Known limitations: Limitations are stated in 'What This Guide Is (and Is Not)', but not detailed, scored 1. License: MIT license is clear, scored 2. Versioning and changelog: No version numbers or changelog, scored 0. Maintenance responsibility: README mentions continuous updates and contribution guide, but does not clearly define maintainer responsibilities, scored 1.
Output usability: Content is well-organized and easy to navigate, scored 2. Marginal value: Provides substantial unique content such as case studies and deep-dive guides, scored 2. Cost-benefit: Free and open-source, but no cost analysis provided, scored 1.
Claim traceability: Some claims (e.g., model prices) have dates but lack specific sources, scored 1. Cross-source corroboration: No external sources are cited, scored 0. Fact-inference separation: Some content distinguishes facts and opinions, but not clearly, scored 1.
- Not found in source: least-privilege scopingGrant only what the task needs: a dedicated account or read-only token, scoped to specific directories and repos.
- Not found in source: confirmation before actingTurn on (or add) a confirmation step before it acts, and try it in a sandbox or test environment before real data.
- Not found in source: data-flow disclosureWatch which external services it contacts (proxy or firewall logs) and keep sensitive data out until you know where it goes.
- Not found in source: sensitive-data handlingUse dedicated, low-privilege, revocable API keys — never production credentials — and keep secrets out of logs.
- Not found in source: dependency securityPin versions and run a dependency audit (npm audit, pip-audit) before installing; prefer running it in a container.
- Not found in source: disclosed external effectsEstablish which external systems it writes to, sends to or changes, and verify with test accounts or repos before production.
- Not found in source: rollback or recovery pathBack up first, or work on a git branch or snapshot, so its changes can be undone.
- Not found in source: verifiable attributionInstall from the official repo or registry and check the publisher and URL to avoid look-alike packages.
- This is a static documentation repository with no executable code or runtime behavior, so any security or execution-related properties cannot be assessed.
- Publisher identity is unverified, and content sources lack detailed provenance; readers should treat specific data (e.g., model prices, benchmarks) with caution.
- The repository has no versioning or changelog; content may change over time, so citations should consider timeliness.