Two acquisitions in 18 months. Three content formats. One publishing workflow. Pick two.
That was the situation when I joined the publishing platform work at Voxy. The company had acquired Nulinga and another platform in quick succession, and the practical consequence was a content authoring environment that had become a negotiation between three different worldviews about what “a course” actually is. One team thought in terms of adaptive lessons. Another was built around live instruction scheduling and session notes. The third had its own content ownership model and publishing pipeline. They were all correct by their own internal logic. None of that logic was compatible.
Why This Was Hard in Ways That Had Nothing to Do with Code
The technical debt was real — the existing platform didn’t support multilingual authoring, which was blocking Voxy’s expansion into Portuguese and Spanish markets. That problem had a relatively clean solution: rebuild with internationalization as a first-class concern.
The harder problem was cultural. The teams from acquired companies had built their workflows around their platforms. They had strong opinions — formed through experience — about how content should be structured, reviewed, and published. When you come in and say “we’re unifying this,” what they hear is “your way of doing things loses.”
I spent a significant amount of time in the first months just listening. Not to generate buy-in (though that mattered too), but because I genuinely didn’t know enough yet. Each team’s workflow had evolved to solve real problems. The goal wasn’t to pick a winner — it was to understand what constraints were fundamental and which were just accumulated habit.
🚧 Need more context: What were the most significant workflow conflicts between the three content teams? Were there specific features or concepts that required the most negotiation to resolve?
What We Built
The rebuilt platform needed to handle three distinct content types without making any of them feel like second-class citizens:
Adaptive lessons — Voxy’s core: AI-driven, vocabulary-dense, personalized to learner context. The authoring flow here needed to support content tagging, difficulty calibration, and multilingual variants.
Live instruction — Sessions between teachers and learners. The publishing side touched scheduling metadata, session materials, teacher notes, and quality standards. Post-acquisition, the different platforms had different ideas about what a “qualified session” meant.
Assessments — Evaluation content that needed its own review and approval logic, separate from lesson content.
Building a unified authoring tool that handled all three required defining shared concepts (what does “publish” mean across types? what does “review” mean?) without flattening the real differences between them.
🚧 Need more context: What were the specific technical architecture decisions? What stack did the rebuilt platform use? What did the design system cover and how long did it take to implement?
The Design System Angle
Running parallel to the platform rebuild, I led the creation of a unified design system. This sounds like a supporting initiative; it wasn’t. The three acquired platforms had three visual languages, three component libraries, and three sets of assumptions about interaction patterns. Instructional designers moving between content types were constantly context-switching in ways that added friction and introduced errors.
The design system created a shared visual grammar. It also dramatically cut design-to-development handoff time — when components are documented and agreed-upon, you stop having the “what did you mean by this design?” conversation on every ticket.
January 2023
I won the Voxy Impressive Performance Award in January 2023. The framing the company used was “exceptional impact during organizational change,” which is accurate but undersells how much of that period was genuinely difficult. Acquisitions are disorienting. People are uncertain about their roles, their teams, and what the company is becoming. Building a platform that multiple teams have to trust required showing up consistently and earning that trust through work, not position.
🚧 Need more context: What were the publishing productivity metrics after the rebuild? How many active instructional designers used the platform? What specifically happened with live instruction operations after unification — different quality standards, scheduling systems?