Back

Dataset Catalog

One place for developers to find, request, and access K-12 data

Jan to Sep 20260 to 1Systems thinking

Context

Learning Commons is CZI's nonprofit developer platform for K-12 education. Open data was scattered across GitHub, docs, and storage links. Licensed data moved through email and Google Drive folders, and access was tracked in spreadsheets.

Problem

Pulling everything into one catalog meant reconciling two mental models. Developers think in datasets. Our platform grants access through licensing boundaries called partitions, and the two don't map one-to-one. Every screen had to translate between them without overpromising what the system could do.

The IM v.360 card from the catalog, by Illustrative Mathematics, marked both open and gated. Three partitions drop in beneath it: Scope and sequence, open, with a Download button; Instructional materials and Assessment materials, each gated, with a Request access button.

Meanwhile, launch slipped three times, access scope flipped twice in one month, and the request flow had to ship before the authorization service it depends on.

Approach

I organized the catalog the way developers already look for data: category, dataset, subset, then download or request. Partitions stayed behind the scenes. Research showed people could find a dataset by name but couldn't tell whether it was worth downloading, so I rebuilt the detail pages around that question.

Three tilted columns of real Dataset Catalog cards drift slowly past, each with its publisher, description, subject tags, and an open or gated badge: U.S. academic standards, IM v.360, US academic standards crosswalks, Math learning components, English Learners Success Forum Guidelines, Math Coherence Map, Math misconceptions, OpenSciEd, ELA learning components, Carnegie Skills Progressions, and XQ Competencies.

I built a production-like React/TypeScript prototype that replaced Figma as our engineering handoff. I also set up agents that synced meetings, Slack, and docs into a decision log and open questions tracker each day, so the team and the prototype stayed current as scope shifted.

Solution

Developers sign in to browse 13 datasets across five categories, check each one's contents and source, and download open data right away. Licensed data goes through a structured request with a sample agreement and a status timeline. Publishers control access through organization roles, domain verification, and audit logs.

A developer requests access to IM v.360 instructional materials. The card opens into the request, which lays out the steps (submit, publisher licensing review, publisher approval). They name their products and users, move through the organization and contact steps, and land on Request submitted. The card then shows the request as submitted.

Impact

The catalog launched on September 22, giving teams an auditable path to education data. Next, I'm watching whether developers can find and get the right data with less hand-holding from our team.

Back