Capability showcase · BlueberryML framework, in deployment

What the framework does in deployment.

ClassGrade is BlueberryML's framework for sovereign partner deployments — exam authoring and grading, outcome-based reporting, school and student dashboards, voice-first speaking and reading assessment, and a student app, all running on the partner's infrastructure under the partner's brand. This site is the capability surface.

BlueberryML's adaptive learning technology powers Chrysalis, deployed across 300,000+ students and 150+ schools in India through our distribution partner EZ Vidya. Each capability below is in production in that deployment today.

One loop: author, grade, report, aggregate.

Three engines run the partner's assessment surface end to end. Exams are authored and graded against a structured map, results write back into per-student reports written by the framework, and school and student dashboards roll the same evidence into a view leadership can act on.

01ExamBuilder

Author, print, and grade exams in one loop.

ExamBuilder generates an exam to a teacher-chosen section structure, lets the teacher edit at question or section level, prints with QR-coded alignment stamps, and grades scanned student work through an anchor-based pipeline against a structured grading map.

Six question types. Bounding-box grading config with image-level overrides. Partner question bank in the same authoring surface.

02Reports

Per-student PDFs against learning outcomes.

The Report engine aggregates exam performance, app activity, LSR scoring, and teacher input into a per-student PDF with LLM-generated narrative sections written against outcome-level evidence — not generic comments.

Level-aware templates for Nursery, Kindergarten, Primary, and Middle school. Sections, voice, and weighting configurable per school.

03Dashboards

School and student views, refreshed weekly.

School dashboards do HPC-style aggregation across grade and subject. Student dashboards drill to outcome coverage, dimension-level progression, and app activity. A scheduled job rebuilds the cached views every week.

Same evidence the Reports draw on. Trends and app activity surfaced alongside assessment data; ad-hoc regeneration available per school.

LSR is the framework's voice surface — Listening, Speaking, and Reading, measured in real time.

Ten activity templates cover discriminative listening, imitative and responsive speaking, reading fluency, reading for expression, and reading comprehension. Each activity emits a position on the five-dimension manifold the framework uses for English assessment.

CEFR-aligned at the surface, so partner curricula and international assessments map to the same coordinate system underneath.

Five dimensions

Accuracy, Pronunciation, Fluency, Vocabulary, Content. Every activity returns a coordinate in this space; reports and dashboards aggregate over it.

Ten activity types

Across Nursery, PP1, PP2, and Class 1, extending upward. Reading fluency includes WCPM, prosody, and phoneme-level scoring against CMU Dict.

The same framework, under the partner's brand.

The framework ships as deployable artifacts, not a single fixed app. Partners can take the off-the-shelf student app, layer a custom application over the same underneath, or run an entirely partner-branded deployment on their own infrastructure.

Student app

Chrysalis Fresh — React Native, iOS and Android.

Kindergarten through Class 1 content today, extending grade by grade. An adaptive skill-test loop picks the next activity from the student's current position; the evidence flows back into the same Reports and Dashboards surfaces above.

Custom apps

Partner-specific builds, layered on the framework.

Same assessment APIs, voice scoring, grading pipeline, and intelligence layer underneath. Custom navigation, content shape, and pedagogical flow over the top. Framework upgrades flow through without forking the partner's build.

White-label

Partner brand, partner infrastructure, partner data.

Framework images deploy into the partner's cloud. The partner is data controller and data processor under their jurisdictional law; BlueberryML is not in the data path. Curriculum localisation handled at the configuration layer.

Underneath every surface above is one framework.

BlueberryML maintains a Riemannian-manifold model of learning with a five-dimension coordinate system for English assessment, and field-theoretic quantities that drive adaptive logic. Every capability on this page reads from and writes to the same substrate.

The mathematical framing and the company posture sit on blueberryml.com. This site stays on what the framework does in deployment.

Read about BlueberryML
Partnership

Interested in deploying this? Start at the partnership page.

ClassGrade does not have a sales motion. The partnership page on blueberryml.com describes what partners bring, what the framework provides, and how the data architecture works. Serious operators evaluating the pattern start there.

Each deployment is partner-led and partner-branded. The framework is the substrate; the surface the end-customer sees is the partner's.

Go to the partnership page