Spark Studio

Spark Studio is IXL’s first AI product, a creative workspace where teachers generate worksheets, lesson plans, and quizzes from a library of 40+ tools. It launched in November 2024, is complimentary for educators with an IXL license, and is live and used at scale across a platform serving 17M+ students. I designed the brand and foundational component system that shipped with it.

This page tells the story in two parts. First, what shipped and what it did. Then a self-initiated exploration of where the product can go next: a data-grounding layer that ties AI output to each class’s real IXL performance data. The exploration is clearly labeled as a concept. The point is to show how I find a product gap and design the extension to solve it.

Prior to launch, I collaborated with the UI design team to define the platform’s brand voice, establish a comprehensive visual identity, and build the foundational component system for Spark Studio. The product went from concept to launch in a fast, deadline-driven push, roughly three months from first designs to release.

One early decision shaped everything after it: rather than adopting IXL’s existing design system wholesale, we built a custom system that kept IXL’s familiar palette so teachers would trust it on sight, while giving Spark Studio room to feel like a new, creative space. The principles we set: accessible and responsive, trustworthy and transparent, familiar and easy to use, systematic and flexible.

Spark Studio launched publicly on November 6, 2024 and became part of teachers’ weekly prep routine. Product-level results reported after launch:

These are outcomes of the whole team’s work, design, engineering, and product together. My share of it is the part teachers see first and trust or don’t: the brand and the component system every tool is built from.

IXL already tracks real-time classroom performance through its Teacher Analytics dashboards. My proposal doesn’t invent anything new, it builds the technical bridge that lets Spark Studio ingest these existing insights, from class-wide trouble spots to specific skill gaps.

AI in education isn’t just a speed race; it is a responsibility of trust. During a short lunch break, Ms. Jones reviewed a Trouble Spots report and manually grouped students before her next class. Her routine inspired the strategy, exposing a universal reality: teachers operate in highly compressed, high-pressure prep windows. That reality anchored the framework on five core principles.

Before any visual polish, I wireframed low-fidelity flows to settle two decisions: how a teacher should control how much data is visible, and what the new homepage should surface. These drafts became the three-mode switch bar and the centralized dashboard that follow.

A toggle only offered on or off; a dropdown buried the choice. A segmented switch won: it keeps all three states visible and one tap away, which set up the Hidden, Semi-Transparent, and Fully Visible framework.

The same product branches by how much data a teacher wants. Mr. Marcus, Ms. Taylor, and Ms. Chen open the same door and land on a homepage tuned to their appetite, then move through the same four steps to a finished document.

Together these paths preview a data-visibility spectrum, Hidden, Semi-Transparent, and Fully Visible, which the next section formalizes.

The form doesn’t change. The source of truth does. What’s behind it changes everything.

Imagine a worksheet generator completely honest about the data backing it. Instead of forcing a one-size-fits-all output, the system evaluates real-time classroom data and dynamically adapts through three confidence levels.

When we shipped Spark Studio, we moved at a breakneck, deadline-driven pace. My role focused on the visible layers: visual identity, component system, design direction. In the rush, the harder structural work was set aside; we never fully grounded the AI’s suggestions in real student data. At the time it was the right organizational call, but it left a critical gap.

The turning point came when I analyzed post-launch teacher feedback. A consistent note surfaced: teachers loved the visuals, but couldn’t trust AI advice that failed to tie back to their classroom context. It was never a visual problem. The system simply lacked an honest, data-backed reason for its recommendations.

So I pushed deeper, into a self-guided, speculative phase: what would it take to ground the AI in real student data, and how much of that data should a teacher actually see? That inquiry became this framework, Hidden, Semi-Transparent, and Fully Visible.

This work remains a speculative proposal rather than a validated feature. I haven’t tested these prototypes with real educators, nor have I resolved the complex privacy implications of surfacing student data this way. Yet my approach to craft changed for good. I used to think design was the surface a product wears. Now I understand design is the reasoning underneath, the quiet, invisible layer that decides whether someone trusts what they see.