Helix: Shopify's AI-powered mobile migration framework
How Shopify rebuilt 300+ screens natively using LLMs with quality checkpoints
1Migration challenge at scale
Shopify faced the massive task of migrating their largest mobile app with over 300 screens from React Native back to native Swift and Kotlin. After successfully completing Shop app's migration in 12 weeks, they developed Helix to handle the more complex Shopify App. The key challenge was maintaining consistent code quality while leveraging LLMs for automation, as raw AI output often lacks maintainability standards required for production apps.
2Helix's core philosophy
Unlike traditional migration tools that deliver large code chunks for review, Helix implements a checkpoint-based approach. It breaks down screens into small, verifiable components that must pass strict quality gates before progressing. This creates a feedback loop where imperfect attempts are automatically refined rather than requiring complete rework. The system combines architectural guardrails with incremental learning from engineer inputs.
3Checkpoint breakdown system
Helix decomposes screens into sequenced checkpoints of increasing complexity, starting with basic skeletons. Each checkpoint is deliberately small - often just one UI component - to enable focused reviews. This allows engineers to validate architectural decisions early while keeping context manageable. The system maintains shippable code quality at every stage rather than accumulating technical debt for later resolution.
4Four-stage quality gates
Every checkpoint must pass four automated verification gates: behavior tests (validating functionality against reference), UI equivalence (pixel-perfect comparison via Gemini), code review (against documented architecture standards), and engineer approval. Gates are strict and non-overridable - the system automatically iterates until all criteria are met. This ensures each incremental change meets production standards before becoming foundation for next steps.
5AI-powered UI validation
Helix uses Gemini's spatial awareness to detect subtle UI discrepancies like margin errors or font sizing issues that traditional diffing misses. The system captures reference and implementation screenshots in matching states, then generates detailed discrepancy reports with severity ratings. This automated visual review catches design inconsistencies that would normally require human QA, while accounting for framework rendering differences.
6Adversarial code review
Two independent AI reviewers analyze each checkpoint against Shopify's thoroughly documented native architecture standards. They examine both functionality and code structure, with all findings requiring fixes before progression. Reviews are context-isolated to prevent bias, and affected tests rerun after corrections. This dual-reviewer system enforces architectural consistency across the entire codebase.
7Engineer feedback integration
Human input focuses on high-value decisions about product scope and UX rather than line-by-line code review. Engineer feedback gets incorporated into Helix's memory, improving subsequent checkpoints. Early migrations require more oversight, but approved patterns enable increasing autonomy. The system can operate overnight, processing multiple screens in parallel while maintaining all quality gates.
8Beyond migration applications
While built for React Native migration, Helix's framework applies to any code generation task. It can develop new features from design specs using the same checkpoint process, or handle architecture refactors by adjusting gate criteria. Logic-only changes skip UI review, showing the system's adaptability. The core innovation is reliable convergence rather than perfect first attempts.
9Engineering productivity impact
Helix demonstrates how AI can augment (not replace) engineering workflows when given proper tooling. By automating repetitive implementation and verification tasks, it lets engineers focus on architectural decisions and product quality. The system achieves what Shopify describes as 'previously impossible' velocity while maintaining high standards - a model for complex codebase transformations.
Original article: Shopify Engineering — Helix: Shopify's AI-powered mobile migration framework