Built for Schools · Privacy-First AI

Feedback that lands.
Data that stays.

IndiLearn builds AI education tools that live inside your school — not someone else's cloud. Student data never leaves the building. Two products, one architecture, zero compromises on privacy.

🏫
On-site AILocal Mac mini · no cloud dependency
🔒
Zero data exposureMeets strict school privacy requirements
📚
Science of ReadingScience of Reading aligned
🧪
Evidence-basedHattie 0.92 feedback effect size
🖥️

Your school. Your server.

Mac mini running local models. No per-token billing. No data sovereignty risk.

🎤

Phoneme recognition never done before

Proprietary neural engine captures isolated grapheme responses with Australian accent support.

🔁

Feedback that closes the loop

Tracks whether students enact next-steps. The missing link in every other feedback tool.

👩‍🏫

Designed by teachers, for teachers

Built around real classroom pain points. Reduces workload rather than adding to it.

Two tools. One thesis.

Every IndiLearn product targets a specific logistics gap between what research says works and what teachers can actually deliver at scale.

Phonics Reading App

The world's first on-device phoneme assessment for young readers

A structured literacy tool for Prep–Year 2, built on the Queensland scope and sequence. Our proprietary speech recognition neural engine captures children's isolated phoneme production in real time — something no other app does — with full support for Australian accents.

6-stage teaching flow — Read letters → Write letters → Read words → Write words → Read sentences → Write sentences. One thing at a time.

Proprietary phoneme recognition — Novel neural engine trained to capture isolated grapheme-phoneme correspondences with Australian accent variants. Never done before.

PencilKit tracing — Multi-algorithm scoring (pixel coverage, Fréchet distance, Hausdorff) gives zone-specific feedback: "top stroke good, lower curve needs work."

On-site LLM content generation — Decodable words and sentences generated using the child's active graphemes and real-session struggle data. Zero cloud calls.

AU/US accent toggle — Configurable for school context. Wired through speech recognition and TTS simultaneously.

⚙️  In development · Pilot schools Q3 2026
Feedback Platform

Individualised written feedback at classroom scale — without the workload

Quality descriptive feedback has an effect size of 0.92 — one of the largest in educational research. Three things stop it happening at scale: it's physically impossible to deliver individually, students don't always apply it, and teacher capacity to analyse work varies. IndiLearn closes all three gaps.

Rubric-grounded feedback — Teacher uploads a unit PDF. System extracts structured success criteria. Feedback is always anchored to a specific criterion.

Enactment tracking — Did the student use the next-step? The system explicitly checks. "You're working on the same thing. Here's another way to think about it."

Teacher dashboard — Class roll, lesson themes (three most common next-steps across the class), and a one-paragraph recommendation for tomorrow.

Student fingerprint — Across a 10-week unit, each student builds a longitudinal record of recurring strengths, recurring next-steps, and trajectory per criterion.

On-site inference — All OCR, feedback generation, and theme synthesis runs on the school's Mac mini. Not a single student work sample leaves the building.

🚀  Pilot build · Late Q3 2026 target
📱
iPad / Web AppStudent and teacher interfaces
🖥️
School Mac miniOn-site · Apple Silicon · Local LLM
🔒
School network onlyData never leaves the building
🧠
CoachingProvider APIProprietary inference layer · versioned prompts

The school owns the intelligence.

Every IndiLearn tool runs inference on a Mac mini that lives in your server room. No per-token API costs. No student data in someone else's training set. No privacy consent forms for every child.

This isn't a cost-saving compromise — it's the procurement story that cloud-based AI education tools structurally cannot tell. A 600-student school amortises the hardware in months, then pays nothing per query, ever.

  • Single CoachingProvider interface — swap cloud to local without changing a line of application code.
  • Versioned, citable prompts — pedagogical voice is first-class. A school that onboards in March doesn't silently get different feedback in November.
  • Predictable cost — hardware amortised over 3 years, zero per-student variable cost. Budgeting is straightforward.
  • DET QLD compliant — student data never crosses network boundaries. Meets the strictest interpretation of Australian Privacy Principles for minors.
  • Upgradeable — as local models improve, schools upgrade the model weight, not the application. The privacy guarantee never changes.

Built on what works.

IndiLearn tools are grounded in the most replicated findings in educational research. Not buzzwords — actual effect sizes.

0.92

Quality descriptive feedback effect size (Hattie)

One of the largest effect sizes of any pedagogical approach measured across 1,800+ meta-analyses. The gap between this potential and classroom reality is not knowledge — it's logistics. IndiLearn closes the logistics gap.

Hattie, J. Visible Learning (2009) · Hattie & Timperley, Review of Educational Research (2007)
73.9%

Australian teachers with burnout

Nearly three quarters of Australian teachers report moderate-to-high burnout. Workload is the single greatest driver. IndiLearn reduces marking and feedback burden — not adding another platform teachers must manage.

Education Daily AU (2025) · BERA Workforce Attrition Study
$35M

QLD Reading Commitment investment

Queensland's 2023 Reading Commitment backed systematic synthetic phonics with a $35M investment, Year 1 Phonics Check, and Australian Curriculum v9.0 rollout. IndiLearn aligns directly to this commitment.

QLD Dept of Education · QLD Reading Commitment (2023)
0.92

Feedback effect size — the highest-leverage move in teaching

Quality descriptive feedback has one of the largest effect sizes of any pedagogical approach. Three things stop it happening: physically impossible at scale, students don't always apply it, and teacher analysis capacity varies. IndiLearn closes all three gaps.

Hattie, Visible Learning · Hattie & Timperley, Review of Educational Research (2007)
4,100

Secondary teacher shortage by 2025 (federal projection)

The federal government's workforce projection, combined with 70% of teachers reporting unmanageable workloads, means the profession needs tools that multiply teacher capacity — not require more of it.

Dept of Education National Workforce Action Plan (2023)
100%

Data sovereignty

The Australian Framework for Generative AI in Schools identifies student privacy as the key concern for Education Ministers. IndiLearn is the only platform that addresses this structurally — not through policy, through architecture.

Australian Framework for GenAI in Schools, Term 1 2024

Why IndiLearn exists.

The research behind the products. The context behind the decisions.

Privacy

Why student data must stay on-site: the case against cloud AI in schools

Australian children's data is uniquely valuable to bad actors. Here's why on-site inference isn't a feature — it's the only responsible architecture for school AI tools.

Workforce

The teacher burnout crisis: 73.9% burnout rates and what technology can actually fix

Nearly three quarters of Australian teachers are burning out. The cause is specific, measurable workload tasks. Technology can reduce them — but only if it's designed to.

AI & Teaching

AI is a tool, not a teacher: what schools actually need from technology

The AI-replaces-teachers conversation misunderstands what schools do. Teachers cannot be replaced. But their logistics gap can be closed.

Evidence

The Science of Reading, Hattie's feedback, and cognitive load: the three pillars behind IndiLearn

Every IndiLearn design decision maps to a specific, replicated finding in educational research. Here's the evidence behind the products.

School Autonomy

School-based decision making: why edtech that dictates practice is failing schools

Commercial programs that require scripted delivery de-skill teachers and ignore local context. IndiLearn is built to work with schools, not tell them what to do.

Technology

The phonics recognition problem nobody solved: how IndiLearn built a proprietary neural engine

Capturing isolated grapheme-phoneme production from young readers is technically unsolved. Here's what we built and why it matters for any classroom.

Architecture

On-premise AI for schools: data sovereignty, predictable cost, and the Mac mini case study

A 600-student school amortises local AI hardware in months, then pays nothing per query. Here's the architecture decision that changes the economics of school AI.

Built for your classroom.
Not for someone else's cloud.

IndiLearn is in active development. Register your school's interest to join a pilot or receive updates as we approach launch.

No spam. No commitment. Your data stays with us — never shared, never sold.