Launching today
Studeia
LMS with AI tutor per course and more
5 followers
LMS with AI tutor per course and more
5 followers
Learning Management System (LMS) / Virtual Learning Environment (VLE) platform with AI tutor per course, native gamification, and structured gradebook — the modern alternative to Moodle for schools, universities, and preparatory courses.
Studeia — AI-Native Learning Platform for Institutions
Studeia is a complete, multi-tenant platform for institutions — schools, universities, exam-prep centers, and companies — to build and run online courses with Artificial Intelligence at the core of the learning experience.
Each institution is a tenant with full data isolation, white-label branding, custom domain, and independent governance. Through a complete educational CMS, institutions build courses with their own material, manage class groups, hold live classes, assess students, and give every learner an AI tutor powered exclusively by the institution's own content.
It's a modern LMS/VLE whose key difference isn't "AI bolted on" — it was designed around AI from the ground up, with governance, pedagogical traceability, and infrastructure resilience as its pillars.
Key differentiators
1. LLM-agnostic (real multi-provider, with automatic fallback)
Studeia doesn't depend on a single AI vendor. The whole platform runs on an intelligent fallback chain:
Each call is routed to the right provider and model for the task (objective grading uses a fast, cheap model; planning and diagnosis use deep-reasoning models).
A circuit breaker monitors each provider's health: if one goes down, traffic jumps straight to the next — no timeout, invisible to the user.
Every AI agent on the platform (tutor, evaluator, content generator, course agent, gamification agent, course reviewer, supervisor) works with any of the four providers — including tool calling. No agent is locked to one vendor's native API.
Cost is metered in real time, tracking input and output tokens separately, applying margins and USD/BRL FX, never using hardcoded prices.
Bring-your-own API keys per institution — each tenant connects its own AI account (keys encrypted with AES-256-GCM). AI cost goes to the institution, fully transparent.
Per-task provider and model choice — the admin decides which provider and model each AI task uses, without breaking anything.
Per-course AI tutor toggle — turn the tutor on or off course by course.
Tenant-customizable prompts — the institution shapes the tutor's behavior and tone.
Strictly institutional RAG — in tenantOnly mode, the tutor answers only from the institution's own material, never leaking the model's generic knowledge. It always cites the source: curated content vs. AI-generated.
AI Supervisor agent — a background moderator analyzes every tutor-chat turn, classifies risk by level and category (inappropriate language, jailbreak, off-topic, etc.) and applies graduated measures. Self-harm is treated as a crisis, not a violation: the tutor is interrupted with a supportive message and resources, and the admin is urgently notified — never as punishment.
Per-plan limits — messages/day, allowed models, and resource caps are configurable and enforced at the metering layer.
BigBlueButton (iframe embed), Zoom, Microsoft Teams, and Google Meet.
For each:
Automatic room creation, attendance tracking, and a provider-resolution cascade (class → tenant default → legacy BBB → environment).
Automatic recording — captured via webhooks (Zoom, Teams) and via a polling cron as a reliable fallback, since webhooks are inherently unreliable.
Native-first transcription — the platform uses the provider's native transcript first (free, normalized from VTT to text), falling back to external STT (Whisper → Google STT) only when needed.
Transcripts become knowledge — a reviewed, approved transcript can be ingested into the course RAG, feeding the AI tutor with what was said in live classes.
OAuth credentials and tokens encrypted (AES-256-GCM); recording access is always server-mediated (the master key never reaches the browser).
3-state lifecycle: active → resolving → resolved, with transitions driven by real evidence (mastery gain, tutor intervention, quiz pass).
Full traceability: each misconception records its source (chat, quiz, interactive, inline quiz), origin session, lesson, triggering question, the student's answer, and linked concepts.
The tutor is misconception-aware — they're injected into the prompt and the tutor addresses them proactively, even generating inline quizzes in the chat to verify resolution.
Bayesian concept mastery and episodic memory (what worked for each student) feed a unified student model.
Multi-dimensional risk assessment across 6 weighted factors (inactivity, performance trend, concept mastery, quiz failure, engagement, active misconceptions) — not simplistic "low progress" detection.
Per-course reports for teachers, coordinators, and the parent portal, with quiz performance, study time, learning profile, class-level aggregated misconceptions, and AI-generated narrative.
Integrates with HeyGen (LiveAvatar) and D-ID, BYO-key model (the institution connects its own account, key encrypted).
Per-course avatar config — provider, avatar, voice, quality, tutor name and personality.
Single-provider sync: the same provider handles TTS and video, eliminating speech-image desync.
Graceful degradation: full avatar → audio (TTS) → text — if the avatar fails, the student is never left without a tutor.
Monthly per-institution quota checked before each session (fail-closed); the master key stays server-side.
Works on web and mobile (via a WebView that reuses the same component).
AutoCourseGenerator — from a document (PDF, DOCX, PPTX), the AI extracts, structures, and generates a full course with modules and lessons in a multi-phase pipeline.
Interactive Lesson Builder with 8 lesson editors (rich text, slides, quiz, video, PDF, assignment, live class, external link).
Course Review Agent — analyzes a course using real performance data (concept mastery, common misconceptions, RAG coverage) and suggests prioritized improvements: add lesson, edit, add exercise, reorder, reinforce.
Gamification Agent — a conversational agent with 19 tools that builds and manages a course's entire gamification strategy via natural chat (badges, contests, rewards, XP, levels).