The largest collection of 500+ real-world Generative AI & LLM system design case studies from 130+ companies. Learn how industry leaders design, deploy, and optimize large language models and generative AI systems in production.
First published: June 14, 2023. Last updated: March 08, 2025
📍 Original Source Attribution: The foundational 500 ML case studies in this repository were originally curated by Evidently AI. This repository builds upon their excellent work, expanding with additional GenAI and LLM-focused content and organization.
- What's Inside
- Featured LLM Case Studies
- Browse by Industry
- Browse by Use Case
- Browse by Company
- GenAI Architectures
- Contributing
This repository documents how companies build and deploy production-grade Generative AI and LLM systems, focusing on:
- Architecture decisions for RAG, fine-tuning, and multi-modal systems
- Scaling strategies for billion-parameter models
- Optimization techniques for latency, cost, and performance
- Evaluation frameworks for LLM outputs and hallucination mitigation
- Deployment patterns across industries
Perfect for:
- AI/ML Engineers implementing LLM-powered features
- Engineering teams designing scalable GenAI architectures
- Leaders planning generative AI initiatives
- Technical interviews on LLM system design
- Ramp: From RAG to Richness: How Ramp Revamped Industry Classification - Enterprise RAG implementation
- GitLab: Developing GitLab Duo: How we validate and test AI models at scale - Testing LLM quality at scale
- Picnic: Enhancing Search Retrieval with Large Language Models - LLM-powered search
- Slack: How We Built Slack AI To Be Secure and Private - Enterprise LLM security
- Discord: Developing rapidly with Generative AI - Generative AI platform
- GoDaddy: LLM From the Trenches: 10 Lessons Learned Operationalizing Models - LLM production lessons
- Tech (90 case studies) - 24 LLM case studies
- E-commerce and retail (119 case studies) - 21 GenAI case studies
- Media and streaming (44 case studies) - 18 LLM case studies
- Social platforms (57 case studies) - 15 GenAI case studies
- Fintech and banking (31 case studies) - 12 LLM implementations
- Delivery and mobility (108 case studies) - 10 GenAI applications
- LLM implementation (92 case studies)
- Generative AI applications (98 case studies)
- RAG systems (42 case studies)
- LLM-powered search (60 case studies)
- NLP & text processing (48 case studies)
- LLM evaluation (36 case studies)
- Fine-tuning approaches (22 case studies)
- LLM inference optimization (19 case studies)
- Multi-modal systems (17 case studies)
- Content personalization (15 case studies)
- OpenAI (8 case studies)
- Anthropic (7 case studies)
- Microsoft (16 case studies)
- Google (15 case studies)
- Meta (12 case studies)
- Hugging Face (9 case studies)
- Netflix (14 case studies)
- LinkedIn (19 case studies)
- GitHub (7 case studies)
- Spotify (10 case studies)
-
Pattern 1: Direct LLM Integration
- Cost-effective for simple use cases
- Examples: GitHub Copilot
-
Pattern 2: RAG (Retrieval-Augmented Generation)
- Improves accuracy with domain-specific knowledge
- Examples: Ramp's Industry Classification
-
Pattern 3: Multi-Agent Systems
- Complex reasoning through agent collaboration
- Examples: AutoGPT-like architectures
-
Pattern 4: Human-in-the-Loop
- Critical applications requiring human oversight
- Examples: Content moderation systems
- 2023 Q1-Q2: First wave of RAG implementations
- 2023 Q3-Q4: Fine-tuning becomes mainstream
- 2024 Q1-Q2: Agent architectures emerge
- 2024 Q3-Q4: Multi-modal systems gain traction
- 2025 Q1: Real-time personalization with LLMs
┌─────────────────┐ ┌─────────────────┐ ┌─────────────────┐
│ │ │ │ │ │
│ Document │────▶│ Vector │ │ │
│ Corpus │ │ Database │────▶│ │
│ │ │ │ │ LLM │
└─────────────────┘ └─────────────────┘ │ Generation │
│ │
┌─────────────────┐ ┌─────────────────┐ │ │
│ │ │ │ │ │
│ User │────▶│ Query │────▶│ │
│ Query │ │ Processing │ │ │
│ │ │ │ │ │
└─────────────────┘ └─────────────────┘ └─────────────────┘
┌─────────────────┐ ┌─────────────────┐ ┌─────────────────┐
│ │ │ │ │ │
│ Base LLM │────▶│ Fine-tuning │────▶│ Specialized │
│ Model │ │ Pipeline │ │ Model │
│ │ │ │ │ │
└─────────────────┘ └─────────────────┘ └─────────────────┘
▲
┌─────────────────┐ │
│ │ │
│ Company │───────────┘
│ Data │
│ │
└─────────────────┘
┌─────────────────┐ ┌─────────────────┐
│ │ │ │
│ Real-time │────▶│ Feature │
│ Data │ │ Computation │
│ │ │ │ ┌─────────────────┐
└─────────────────┘ └─────────────────┘ │ │
│ │ │
┌─────────────────┐ ▼ │ │
│ │ ┌─────────────────┐ │ │
│ Batch │────▶│ Feature │────▶│ LLM │
│ Data │ │ Store │ │ Application │
│ │ │ │ │ │
└─────────────────┘ └─────────────────┘ └─────────────────┘
Contributions are welcome! Help us document the evolving GenAI landscape:
- Fork the repository
- Create a new branch
- Add your LLM/GenAI case study using the established format
- Submit a pull request
See CONTRIBUTING.md for detailed guidelines.
This repository is licensed under the MIT License - see the LICENSE file for details.
- Thanks to all the companies and engineers who shared their LLM/GenAI implementation experiences
- All original sources are linked in each case study
⭐ Found this valuable for your GenAI/LLM work? Star the repository to help others discover it! ⭐
