He helped grow Google Cloud more than 10x, to over $12B, as its President. Now at Sonar, he makes sure AI-written code gets checked before it ships.
Session
Main Stage
Date
Monday, November 30
Level
Everyone
// AI Engineering Skill /
Making systems secure and reliable
// 01 / Bio
Tariq Shaukat is CEO of Sonar, which works with more than half of the Fortune 500 and millions of developers to help them run agentic software factories with confidence: saving tokens, raising quality and keeping code secure. He has helped scale the company to $500M in ARR while keeping a Rule of 60+ operating profile, and lifted its growth rate by more than 70%.
Before Sonar, Tariq was President of Bumble, where he helped lead its IPO and scaled the business profitably past $1B in revenue. Earlier, as President of Google Cloud, he built its go-to-market operations, led product and engineering for its industry and AI solutions, and helped grow the business more than 10x to over $12B in revenue. He has also held senior roles at Caesars Entertainment, McKinsey & Company, Trilogy Software and The McKenna Group.
He serves on the board of directors of Gap Inc. and the board of trustees of Public Storage. He holds a B.S. in mechanical engineering and an M.S. in technology and policy from MIT, and an M.S. in mechanical engineering from Stanford. He lives in Austin with his wife, Gigi, their two sons and their dog, Midas.
// 02 / The session
What agent-written code looks like at Fortune 500 scale
Abstract
AI agents now write most of the code many teams commit, and much of it is good. Yet the gains stall downstream. Pull requests pile up, review becomes the bottleneck, and releases lag far behind code creation. Teams now have more code than they can confidently ship.
Tariq covers what breaks when agents run at scale, drawing on production data and recent research: errors that compound over long-running tasks, agents that lose the thread in large codebases, LLM judges that return different verdicts on identical inputs, and human reviewers overwhelmed by the volume of code agents produce.
For each failure mode, he’ll share the engineering fix, then show how the fixes combine into the code confidence formula: intent, guiderails, verification and evidence, applied on every change, at any volume, at agent speed. The result is a practical framework for building software factories that scale.
What you’ll learn
Where agent-written code breaks at scale, from compounding errors to overwhelmed reviewers
The engineering fix for each failure mode
How intent, guiderails, verification and evidence combine into one framework
AI DEV · BY DEEPLEARNING.AI
Join us at AI Dev
Meet the engineers shipping AI to production, and leave ready to do it yourself.