From Google’s Go team to a production marketplace: Inside Dylan Le’s engineering bet

Opinions expressed by Digital Journal contributors are their own.

Le designed and shipped four production systems before Licensed To Glow had its first engineering hire. The decision to treat partner integration as a first-class system, not an afterthought, is the bet he says will compound over the next several years.

Before he co-founded Licensed To Glow, Dylan Le’s last full-time engineering role was at Google, on the team responsible for the Go programming language. The Go team is small, deliberately so, and each engineer on it owns a defined surface area of the language, the runtime, or the toolchain. Code shipped from the team flows on a regular release cadence into Google’s production stack and the global open-source Go ecosystem.

“The Go team teaches you how to think about engineering at the level of the substrate,” Le said. “Whatever you ship is something that other engineers depend on. You can never be sloppy with abstractions, because the abstractions you choose become the constraints other people work inside.”

When Le co-founded Licensed To Glow with chief executive officer Odette Yang, the engineering problem in front of him was different in shape but, he argues, similar in structure.

The marketplace as a substrate

Licensed To Glow operates as a two-sided subscription marketplace. Members buy monthly or multi-month subscriptions and book appointments in real time with vetted partner venues in their neighborhood; the company pays each partner at pre-negotiated rates per service rendered. The platform comprises four production systems: a consumer iOS application, a parallel consumer web application, a partner-side venue platform portal, and an internal admin dashboard.

What is technically demanding about the platform is not any of those four systems in isolation. It is the integration.

“Subscription state has to remain consistent with appointment availability across both sides of the market,” Le said. “Payments and identity have to propagate cleanly across mobile, web, and the venue portal. The marketplace has to settle correctly between each member’s recurring plan and each partner’s pre-negotiated rate. And the whole thing has to integrate in real time with each partner’s existing scheduling system, including Square, Boulevard, Zenoti, and many more. None of those is hard on its own. All of them being correct at the same time, at scale, is hard.”

Architectural decisions

Le made the foundational architectural decisions on the data model, the authentication system, the payment integration, the partner synchronization layer, and the deployment pipeline before the company’s first engineering hire. He also authored the majority of the production code base alone.

The decision Le points to as the most consequential was treating the partner integration layer as a first-class system, rather than as an adapter.

“In a typical marketplace startup, you treat the partner side as a thin layer on top of your own database,” he said. “That works until you grow. Then it stops working, and the rebuild eats six months of engineering time. We treated the partner integration as the foundation, not the top layer. The synchronization with each partner’s scheduling system was designed from day one. That decision is the thing that lets us eliminate double-bookings as a structural risk, not as an operations problem we have to fight every week.”

The unified backend that the consumer apps share is, in his framing, downstream of that earlier decision. Members can move between the iOS app and the web app without state loss because the backend was designed once.

From substrate to roadmap

Everything the company is building now sits on the substrate Le shipped. Licensed To Glow has said it plans to extend the platform into payments and consumer financing, private-label and procurement infrastructure for partner venues, and a partner operating-system layer. Each of those extensions, in Le’s account, depends on architectural decisions made before the company had its first paying member.

“You only get to set the foundation once,” Le said. “If you set it well, you spend the next several years adding capability on top of it. If you set it badly, you spend the next several years rewriting it. The engineering bet at the seed stage is really a choice between those two timelines.”

Eight months after launch, the company has reached $2.4 million in gross annualized revenue and operates across eight U.S. metropolitan markets, according to figures shared by the company. Salon partners on the platform have reported to the company up to four-fold increases in monthly visits, with capacity utilization moving from 50 % to 90%.

Engineering leadership at the founding stage

Asked what surprised him about engineering leadership at a startup, Le pointed to the breadth of the role.

“The Google Go team taught me to think deeply about a narrow surface area,” he said. “Building Licensed To Glow taught me to think coherently about a wide one. As a chief technology officer at the founding stage you are doing the architectural design, the production engineering, the technical investor diligence, the integration negotiations with vendors, the technical evaluation of every prospective partner venue, and the engineering hiring process. The thing that has to hold across all of those is consistency of judgment. You cannot be a careful engineer in the morning and a sloppy operator in the afternoon. The role does not allow it.”

His advice for engineers considering the founding-team path is similarly direct.

“Do not start a company because you want to be a chief executive,” he said. “Start one because there is a specific structural problem you understand technically, and you can see the path from a working prototype to a system that compounds. If both of those things are true, build it. If they are not, the company will not survive the gap between the seed stage and the moment the architecture has to scale. That gap eats most early-stage companies. What carries you across it is engineering.”