Search Work Demands Engineers Who Can Cross Competencies
Shopify’s engineering culture places a high value on “range”—the ability to move across areas of expertise. A developer might start in data science infrastructure and later shift to working on Ruby internals. That same principle governs how the search and discovery team is structured. Rather than splitting work between data scientists and engineers, Shopify treats search as a single discipline that requires practitioners to hold both skill sets simultaneously.
Search has historically been treated as a niche specialty, and teams often tried to hire “magical search unicorns” who rarely exist. In practice, this led to siloed groups—data teams and engineering teams working on the same problem without full context. The result was dysfunction, slower production timelines, and unnecessary politics.
The Cost of Siloed Data and Engineering Teams
When data scientists work without engineering context, they tend to chase cutting-edge machine learning research without considering how a model will be delivered to production. They iterate on a new model, then hand it off to engineers to translate, treating deployment as someone else’s problem.
Engineers, on the other hand, lack the data and relevance context needed to make the right tradeoffs. A pure engineer might not know where to tweak a model to reduce bloat without hurting relevance. They can’t make the minute-by-minute decisions required to balance relevance, performance, and stability. Without the full picture, they either implement the model blindly or avoid the complexity altogether—especially when their incentives center on service-level agreements and avoiding on-call incidents.
Each group holds only half the skills needed to make good decisions. Coordination between them adds needless time to production deployment and creates friction.
Range as the Solution
At Shopify, the search team draws few lines between “data” and “engineering” work. Engineers must grow data science skills: they learn to build and run experiments, think scientifically, and evaluate model quality. Data scientists are pushed to become good engineers: they write high-quality, performant, testable code, and take responsibility for getting a model to production and maintaining it.
This matters because search development requires making intricate tradeoffs between correctness, scalability, performance, and maintainability. An elegant solution is the simplest one that satisfies all constraints, and that requires fitting both data science and engineering considerations in one brain.
Management on both sides supports this approach with a commitment to close collaboration. At the individual contributor level, the team doesn’t think of itself as two groups. It’s one team with members reporting to different leads, organizing, planning, and executing together without carving out territorial fiefdoms.
Why Range Will Define Future Engineering Work
The problems of tomorrow will increasingly involve “smart” user interactions. Users will want to talk to systems, start with curated options tailored to their preferences, and be guided through paths without obvious irrelevant options filtered out. These capabilities aren’t add-ons to existing applications—they are increasingly the core of what gets built.
Search and discovery at Shopify is only the beginning. The more personalized and conversational products become, the more engineers must push into data work, and vice versa. The future isn’t specialization within data science and engineering. It’s having the range to move between both.
