Smithsonian Learning Lab 2.0 · EDC · BrightQuery · RLMG

From six million objects to classroom-ready lessons

The Smithsonian's collections are among the deepest teaching resources anywhere, and teachers struggle to put them to work. Our solution pairs a carefully curated library with a practical teaching product: Smithsonian educators and curators decide what publishes, an AI co-designer helps teachers draft classroom-ready lessons from real objects, and every resource is reviewed by people before it reaches a classroom. Built on a foundation you can audit, delivered in two deliberate years.

01

Teachers are short on time and confidence

6M+

objects exposed today

yet the word 'collection' means nothing to a teacher's Tuesday

~5+2

materials a teacher keeps

five supplemental, two core - attention is the scarcest resource

64%

of found resources judged 'not worth using'

volume without vetting reads as noise

#1

complaint: slow loads

and classrooms run on Chromebooks

A teacher with five minutes of preparation needs a small set of resources that are age-appropriate, defensible, engaging, and easy to use - aligned to the standards she is accountable for, and fast on the devices her students hold. Everything in this solution follows from taking that seriously.

02

Our answer, in one sentence

Federate the collections, curate them carefully, let AI draft with guardrails, let people decide, and deliver into the tools classrooms already use.

A curated library

Teachers search a set that Smithsonian educators and curators have chosen and prepared. A readiness review decides which objects can become classroom-worthy, and those enter the library. Deliberately small, deliberately useful.

AI as a co-designer

A teacher states what she teaches - textbook, topics, standards - and the studio drafts object-based lessons for her to shape. The AI works within subject and writing-level guardrails, and its drafts are always reviewed by a person before any classroom sees them.

People decide what publishes

A resource publishes only after a Smithsonian reviewer approves it, and the platform itself enforces that rule. Every step is recorded: who created it, who reviewed it, and when.

03

The architecture

Read it bottom-up: the collections stay where they live; a guardrailed factory prepares candidates; human review is the narrow waist everything must pass; the curated catalog feeds one application of role-scoped workspaces; and lessons flow out to classrooms through connectors and open standards. The qualities on the left rail - accessibility, security, privacy, provenance, performance - are engineered into every layer rather than scoped separately.

508 accessibility · security · privacy and COPPA · provenance · performance - engineered into every layerCLASSROOMS AND CONNECTED SYSTEMSTeachersfind · make · teachStudentsstudent view · no PII wallsGoogle Classroom (Year 2)further connectors on the menuOpen APIspartners embed trusted contentlessons flow out only after human reviewONE APPLICATION - 22 MODULES IN 5 WORKSPACES, SHOWN BY ROLE AND BY SUBSCRIPTIONLIBRARYHomeDiscoverSubjectsMuseumsStandardsSetsSTUDIOLesson StudioMy LessonsCommunityCLASSROOMTeachAssignmentsStudent AppCURATIONReviewSignalsCoverageCorpusPoliciesOPERATIONSInsightsPeopleConnectorsMigrationSettingsTHE CURATED CATALOG - deliberately small, classroom-readytyped resources teachers recognize (lesson · activity · assessment) - standards-verified, rights-aware, provenance-bearingHUMAN REVIEW AND APPROVALSmithsonian reviewers approve each resource before it publishesevery step recorded: who created it, who approved it, whenAI-ASSISTED PREPARATION - guardrailed, always human-reviewedreadiness gating (transform now / enrich / defer) · classification and metadata · guardrailed drafting (subject + writing level) · every output parked for human reviewSMITHSONIAN COLLECTIONS - 21 museums · the National Zoo · 9 research centers · EDAN and Open Access - federated: the artifacts stay where they live

04

A foundation you can audit

We did not start from a feature wishlist. We distilled the solicitation and its environmental scan into 181 contracted capabilities - one row per promise - then planned the product downward from them: 22 application modules, 495 named features, and the sprint backlog beneath. Every feature traces to a capability; every capability traces to a line in the solicitation. Ask about any screen and we can show you the requirement it answers - and the other way around.

One solicitationRFP 26RF0010901 + Environmental Scan181 contracted capabilitiesone row per promise - the language of the contract22 application modulesthe navigation - what each role sees and uses495 planned featurespages, actions, workflows, jobs - the language of the build1,000+ sprint storiesthe working backlog, split at sprint planningtraceable in both directions

05

One application, five workspaces

The product is organized around how schools work. Each workspace is a set of modules that appear for the roles that need them - each role sees only the modules it uses - and modules switch on as the Smithsonian subscribes to them, so the product grows without ever being rebuilt.

LIBRARY · find

teachers, curators

HomeDiscoverSubjectsMuseumsStandardsSets

the vetted catalog: search, browse spines, verified standards, curated sets

STUDIO · create

teachers, SI educators

Lesson StudioMy LessonsCommunity

the co-designer drafts; the teacher shapes; peers share and remix - with credit

CLASSROOM · deliver

teachers, students

TeachAssignmentsStudent App

present mode, send-to-Classroom, and a student view with the scaffolding hidden

CURATION · steward

Smithsonian editors

ReviewSignalsCoverageCorpusPolicies

the approval queue, demand signals, coverage maps, and rules the Smithsonian edits without code

OPERATIONS · run

OET, district IT

InsightsPeopleConnectorsMigrationSettings

dashboards, accounts and roles, connector health, legacy continuity, the module switchboard

And everywhere

508/WCAG accessibility, security, privacy and COPPA, provenance, and Chromebook-lean performance ride every module as acceptance criteria. They are qualities of the product, not line items on a menu.

06

Standards of care, built in

The Smithsonian's name carries real weight with teachers; these are the practices that protect it.

Human review, always

No AI-drafted content reaches a teacher or student without a person's recorded approval. Teachers create, curators review, the Smithsonian decides - and the platform keeps the record for every item.

Standards that mean something

An alignment badge appears only when a standard is verified against a real framework registry and confirmed by an educator. Resources claim only the standards they genuinely support.

A complete record

Every resource can answer: which object am I from, who transformed me, who approved me, and when. Human and AI contributions are attributed alike, exportable in open formats.

Privacy before reach

Consent-gated analytics that never deny content, minimal student data by design, COPPA posture cleared before a single classroom joins the beta.

07

Two deliberate years

Year 1 specifies and validates: Smithsonian-approved requirements, alpha prototypes tested with real teachers, then the Functional Prototype Platform on real collection data. Year 2 puts it in classrooms - the beta cohort, two subjects, four grade bands, 24 exemplar lessons, one integration done well - then tests, launches, and reports. Between them sits an honest gate: the Smithsonian decides, on prototype evidence, what grows.

YEAR 1 · SPECIFY AND VALIDATErequirements, functionality, andtechnical specifications, SI-approvedcontent audit + selection rubricalpha prototypes tested with teachersYEAR 1 · BUILD THE PROTOTYPEthe Functional Prototype Platform:modular APIs, educator-centereddiscovery and authoring, analyticseducator sessions on real dataYEAR 2 · BETA, CONTENT, LAUNCHclassroom beta with real teachers24 exemplar lessons + templatesGoogle Classroom connectedfull testing package, then launchgate: specifications approvedgate: prototype delivered + renewalcheckpoints where the Smithsonian decides, on evidence, what grows next

08

Start focused, grow deliberately

Because the product is modular at the navigation level, growth is a series of small, confident decisions instead of one giant bet. Each module and each connector can be adopted on its own: add Canvas when districts ask, add learner supports for multilingual readers when the program funds it, run migration waves when the legacy agreement lands. The prototype is the first configuration of the same product; the Year-2 beta and the launch run on it.

The growth path, concretely

  • Year 1 turns on the prototype: the Library, Lesson Studio, Standards, the curation workbench, and baseline insights - everything the alpha and prototype sessions need.
  • Year 2 turns on the classroom: the beta cohort, the 24 lessons, the Google Classroom connector, the full testing package, and launch.
  • Beyond: further connectors, the Student App at scale, community sharing, learner supports, migration and legacy continuity, state standards packs - each an increment, none a rebuild.

09

Four hands on one product

Education leads it, engineering powers it, design shapes it, and the Smithsonian governs it - a division of labor where every party does what it is best at.

Learning Lab 2.0one product, four hands on itEDCprime - pedagogy, research,the teacher cohort, programBrightQueryAI and platform engineering -backend, APIs, pipelines, integrationsRLMGthe educator-facing experience -interface design and front endSmithsonianthe collections and the data -editorial approval, privacy, security