Shopify CTO Mikhail Parakhin details the company's rapid "AI Phase Transition," showcasing near 100% internal AI tool adoption and an unlimited token budget for employees. He highlights the impact of internal platforms like Tangle for ML experimentation, Tangent for democratized auto-research, and SimGym for customer simulation, all leveraging historical data to drive significant productivity and performance gains across the e-commerce giant.
Summarized by Podsumo
Near 100% AI Tool Adoption: Shopify employees are rapidly adopting AI tools, with an unlimited token budget and a preference for advanced models like Opus 4.6+.
Tangle & Tangent for ML Ops: Tangle provides a collaborative, efficient (content-hashed caching) platform for ML experiments, while Tangent enables auto-research loops that democratize optimization for PMs and engineers, yielding dramatic improvements in areas like search throughput and storage reduction.
SimGym for Customer Simulation: Shopify's SimGym uses decades of historical customer data to train AI agents for realistic customer simulations, allowing for counterfactual analysis and proactive merchant recommendations with high correlation to real-world outcomes.
AI-Driven Code Quality: The explosion of AI-generated code necessitates rigorous PR review processes and critique loops with high-quality models to maintain code quality and prevent increased bugs in production.
Liquid Neural Networks in Production: Shopify is actively using Liquid Neural Networks, a non-transformer architecture, for super low-latency search (under 30 milliseconds) and long-context offline categorization, demonstrating its competitiveness for specific use cases.
"We effectively fund unlimited tokens for everybody. We do try to control the models that people use, but from the bottom, not from top. Like, we basically say, hey, please don't use anything less than Opus 4.6."
— Mikhail Parakhin
"If your store had previous customers, rather than it's a new store... it helps tremendously. In just correlation in forecast, yeah, we take your previous customer's behavior and we create agents that duplicate those specific distribution of customers that you get."
— Mikhail Parakhin
"Liquid's been taking quite a bit of at least internal Shopify share. And the reason I'm excited is yeah, because it's the only non-transformer architecture that I found being genuinely competitive."
— Mikhail Parakhin