Empowering Enterprises with PedOS 1.1 Lumina: The LearnAdapt Ecosystem
As educational institutions and corporate learning environments scale, the demand for personalized, highly governable AI accelerates. However, deploying AI in education requires more than just calling an LLM API—it requires strict pedagogical orchestration, safety guardrails, and multimodal capabilities.
Enter the LearnAdapt Platform, powered by our latest operating system runtime: PedOS 1.1 Lumina.
LearnAdapt is not just a single application; it is a comprehensive infrastructure designed for enterprise-scale educational deployments. With the LearnAdapt SDK and PedOS at its core, organizations can rapidly deploy, govern, and scale domain-specific AI tutors.
Here is a look at the core infrastructure and the flagship applications currently operating on the platform.
The Core: PedOS 1.1 Lumina & The LearnAdapt SDK
At the heart of the platform is PedOS 1.1 Lumina, our proprietary agentic orchestration engine. PedOS acts as the strict gateway between student inputs and raw AI models, ensuring that every interaction is safe, Socratic, and pedagogically sound.
- Governability & Safety: Strict system prompts, interaction logging, and real-time moderation prevent hallucinations and off-topic interactions.
- Multimodal capabilities: Native support for spatial reasoning, full-duplex voice streams, and visual canvas analysis.
- LearnAdapt SDK: A powerful development kit that allows enterprise partners to build custom AI educational tools in days, not months. The SDK provides drop-in React components for voice, canvas, and chat, automatically wired to the PedOS runtime.
The Ecosystem: Purpose-Built AI Applications
LearnAdapt’s architecture allows multiple tenant applications to run on the same unified infrastructure. Our flagship suite demonstrates the flexibility of the PedOS runtime across entirely different educational domains.
Related research
- From Tools to Teacher-Built Teammates — no-code pedagogical plugin authoring and governed deployment with LearnAdapt Agentic Studio and PedOS. arXiv · DOI
- Beyond Prompt-to-App — accountable translation in teacher-facing agentic authoring. Preprint; not peer reviewed. arXiv · DOI
- EduPluginBench — executable assurance for AI-generated educational plugins. arXiv · DOI
- Auditable Release Control for Pedagogical Leakage in LLM Tutors — an authorization-aware release boundary evaluated under declared contracts. arXiv · DOI
- ES-LLMs architecture for adaptive tutoring — accepted AIED 2026 full paper. arXiv · Published DOI
🧮 Archie Math: Real-Time Conversational Tutoring
Archie Math transforms quantitative education by providing a live, conversational AI tutor. Archie doesn’t just give answers; he listens, speaks, and solves with the student.
- Full-Duplex Voice: Students can interrupt Archie, think out loud, and receive real-time verbal guidance.
- Atlas Knowledge Graph: Every interaction is mapped to a massive, empirically validated skills graph (featuring over 1,200 skills and 4,800 pedagogical citations), ensuring targeted remediation.
Related research context
RankCert studies when a simulated benchmark should certify or abstain from selecting among AI-tutoring policies under structural uncertainty. It is methodological context, not an evaluation of Archie Math. arXiv · DOI
🗣️ Calliope: Conversational Assessment for Learning
Calliope Speak tackles the complex challenge of assessing oral communication and language proficiency.
- Formative Feedback: Calliope evaluates student audio submissions against deep rubrics, providing actionable feedback on accuracy, depth, and clarity.
- Human-in-the-Loop: Designed for educators, Calliope seamlessly integrates AI scoring with human justification, ensuring teachers retain ultimate authority over evaluations.
Related research
CALLIOPE: A Source-Grounded Oral Assessment System and Synthetic Readiness Evaluation — technical report based on synthetic fixtures and software checks; no human-participant outcomes or scoring-validity claim. arXiv · SSRN · DOI
🎨 Critsly: AI Critique Infrastructure for Design
Critsly brings AI into the design studio. Design education relies on spatial reasoning and visual feedback, areas where traditional text-based AI fails.
- Spatial Telemetry: Critsly analyzes student canvases, sticky notes, and diagrams, understanding the geometric and semantic relationships between elements.
- Constructive Critique: The AI acts as a design coach, pushing students to iterate on their ideas rather than solving the design problem for them.
Related research
- Critsly and StudioCrit — simulation-based readiness evidence for an artefact-aware critique workspace and architecture-studio research mode. arXiv · SSRN · DOI
- TeamLens in Critsly — consent-based team-composition interface evaluated with synthetic fixtures and software tests. arXiv · SSRN · DOI
- Critsly: An Artefact-Aware AI Critique Teammate for Design Education and Project-Based Learning — accepted AIED 2026 Interactive Events contribution. arXiv · DOI
📖 Clio: Generative Publishing Studio
Clio is a multimodal generative publishing platform designed to autonomously research, write, format, and compile highly personalized educational assets and textbooks.
- Generative Assembly: Autonomously drafts and compiles high-fidelity, pedagogically structured textbooks, curriculum frameworks, and study guides (worksheets, slides, notes) tailored to specific student profiles or learning objectives.
- Cognitive Verification: Leverages safe Cognitive Engines and Semantic Reasoners to systematically cross-reference and verify all compiled content, preventing factual errors while maintaining strict academic and safety compliance.
🧭 LearnAdapt Praxis: Adult Workplace Design·AI Learning
LearnAdapt Praxis helps adults turn workplace challenges into AI-supported services they can explain, test and improve. Its five-stage journey—Frame, Learn, Build, Validate and Transfer—connects supported project work to a separately controlled transfer task and facilitator review.
- Inspectable learning journey: Keeps the initial response, versioned specifications, coaching records, validation observations, separate transfer response and facilitator feedback together for review.
- Controlled transfer: Praxis can withdraw its own coaching, earlier-work access and export during a new task, creating a clearer in-application condition while recognising that no software can establish whether external assistance was used.
- Human-centred ecosystem: Praxis uses shared LearnAdapt account and application-runtime foundations while retaining its own workflow and services within the wider Agentic Studio and PedOS programme.
Related research
The LearnAdapt Praxis technical report evaluates selected workflow, authority, recovery and provenance properties using synthetic project episodes. It does not establish learning gains, model-output quality, unaided-assessment validity, population usability or production capacity. SSRN · DOI
Bring LearnAdapt to Your Enterprise
Whether you are a university looking to scale personalized tutoring, or a corporation needing to train thousands of employees with high-fidelity assessments, LearnAdapt provides the bedrock.
With PedOS 1.1 Lumina, the days of unreliable, unmonitored AI chatbots in education are over. We are building the operating system for the future of learning.
Ready to explore what LearnAdapt can build for you? Contact our enterprise team today.