

CodeRabbit and RAG for Code Review with Harjot Gill
40 snips Jun 24, 2025
Harjot Gill, Founder and CEO of CodeRabbit, explores the transformative potential of generative AI in code reviews. He discusses how AI can enhance code quality and security seamlessly within tools like GitHub and VS Code. The conversation dives into the architecture of CodeRabbit, the challenges of managing context in AI models, and the importance of transparency in AI-generated insights. Harjot also shares insights on building a tech startup, strategies for community engagement, and the future implications of AI in software development.
AI Snips
Chapters
Transcript
Episode notes
CodeRabbit's Growth Story
- CodeRabbit started as an internal tool and grew independently into an AI code review startup.
- It now has over 100,000 daily users and spans various industries.
Deep Reasoning with Sandboxed Agents
- Code review requires deep reasoning and can tolerate latency unlike code generation.
- CodeRabbit uses sandboxed environments where AI agents run CLI commands to dynamically explore code.
Optimize Multi-Model Pipelines
- Use an ensemble of multiple models optimized for different pipeline stages.
- Employ cheaper, faster models for summarization and expensive reasoning models sparingly for cost efficiency.