Coinbase Transforms Its Engineering with AI and 1,000 Engineers

A Transformation Orchestrated by Chintan Turakhia

Chintan Turakhia, Senior Engineering Director at Coinbase, has orchestrated a major transformation within the organization. With over 1,000 engineers under his leadership, he has embarked on integrating large-scale artificial intelligence tools. This initiative aimed to rewrite Coinbase's self-custody portfolio to transform it into a consumer-facing social application, all within an ambitious timeframe of six to nine months. AI has been leveraged to multiply the team's capabilities, resulting in impressive efficiency gains. For instance, the time required to review pull requests was drastically reduced from 150 hours to just 15 hours. Additionally, the cycle from user feedback to delivered features has been significantly compressed.

Strategies for AI Adoption

Adopting AI in a large engineering organization like Coinbase requires well-defined strategies. Among these, the "speed run" technique has proven particularly effective. It enabled a team of 100 engineers to submit 70 pull requests in just 15 minutes. This method relies on identifying and replicating the behaviors of advanced AI users. To encourage the adoption of these tools, it is crucial for engineering leaders to get directly involved. Their hands-on engagement with AI tools serves as a model and encourages teams to follow their example.

Building Custom AI Agents

Coinbase has also implemented custom AI agents that seamlessly integrate into existing workflows. These agents are designed to optimize internal processes and enhance productivity. The metrics used to measure the impact of AI on engineering velocity are essential for evaluating the success of these initiatives. They allow for quantifying improvements and adjusting strategies accordingly.

Technological Tools for Innovation

To successfully carry out this transformation, Coinbase has relied on a series of technological tools. Among them, Cursor, Linear, Slack, ChatGPT, Claude, and GitHub Copilot have played a key role. These tools have enabled the analysis of AI adoption patterns and facilitated communication and collaboration within teams.

Chintan Turakhia's Online Presence

For those wishing to follow Chintan Turakhia's progress, he can be found on LinkedIn and X.

Demonstrations and Discussions

In a detailed episode, several aspects of this transformation were discussed. Chintan's introduction was followed by a discussion on rewriting the application with the help of AI. The importance of leadership conviction and practical demonstration was emphasized. The "PR speed run" technique was explained in detail, showcasing how it transformed team adoption. Demonstrations illustrated the implementation of real-time feedback to features, the use of Cursor to analyze AI adoption, and the construction of a live feedback capture system using AI transcription.

Automation and Practical Tips

The use of custom Slack bots has allowed for the automation of engineering workflows, making processes smoother and more efficient. Tips were shared to promote AI adoption within organizations, highlighting the importance of leader involvement and aligning tools with team needs. A personal use case was presented, illustrating how AI can be used for wine selection based on taste preferences.

Final Thoughts

The episode concluded with rapid-fire questions and final reflections, providing a recap of the key points discussed and thanking participants for their contributions to this enriching conversation.