About the Platform
Knowledge Without Gates
NaySahul University is an open-source, multi-agent AI academic platform for real Engineering, Science, and Technology education. It is not an accredited institution. Its AI content is subject to human review.
Mission
NaySahul University is designed to provide access to genuinely rigorous undergraduate and graduate-level instruction in Engineering, Science, Computing, and Mathematics — without tuition barriers, institutional gatekeeping, or geographic limitations.
The platform does not claim to replace accredited institutions or award recognized credentials. It provides an AI-powered learning environment with transparent academic standards, open educational resources, and human oversight for consequential decisions.
Our goal is to demonstrate that open-source, multi-agent AI can deliver academically structured, factually grounded, and pedagogically sound instruction at scale — while being honest about the limitations of AI in education.
How AI Teaching Works
Teaching is delivered by a controlled multi-agent system. Each agent has a clearly defined role, an allowlist of tools, resource limits, execution traces, and evaluation tests. The backend — not the model prompt alone — enforces security and authorization.
Academic Planning Agent
Understands student goals, identifies prerequisite gaps, recommends study sequences.
Curriculum Architect Agent
Builds structured programs, courses, modules, and measurable learning outcomes.
AI Professor Agent
Delivers lessons, answers questions, uses examples, derivations, and asks knowledge-check questions.
Teaching Assistant Agent
Identifies misconceptions, provides graduated hints, generates targeted practice.
Laboratory Instructor Agent
Manages isolated lab environments for coding, simulations, and reproducible research.
Assessment Designer Agent
Generates rubric-aligned questions, assignments, and assessments calibrated to course objectives.
Grading & Feedback Agent
Evaluates work against explicit rubrics, explains deductions, supports human review.
Research Mentor Agent
Helps graduate students formulate questions, find literature, and track citations — never fabricates papers.
Academic QA Agent
Verifies alignment, factual consistency, citations, accessibility, and prevents publishing when checks fail.
Student Support Agent
Answers policy questions from approved data, escalates consequential decisions to staff.
Experience a live AI-powered lesson — no account required to preview. Powered by Ollama running on dedicated inference hardware.
Architecture
The platform uses a Next.js frontend (TypeScript, Tailwind CSS) with a Python/FastAPI backend (planned Phase 2+).
Academic data is stored in PostgreSQL with pgvector for semantic retrieval. Redis handles caching and background tasks. Object storage holds permitted learning assets.
AI teaching is powered by a provider-independent model gateway supporting Ollama (local), and pluggable cloud providers. All model calls are logged and auditable.
The multi-agent orchestrator coordinates agents through an explicit state machine. Agents cannot invoke one another indefinitely or execute arbitrary tools. Every agent action produces a structured trace and audit event.
Student code runs only in isolated, resource-limited sandboxes with no access to host secrets, production networks, or other tenant data.
Responsible AI Policy
Limitations Disclosure
AI-generated academic content may contain errors. No AI professor is correct 100% of the time. All published course content passes automated checks and human review before student access.
- Citations are displayed for all factual claims where sources support the answer.
- AI professors disclose uncertainty and missing evidence explicitly.
- Grade changes and credential issuance require human-authorized workflows.
- Research agents never fabricate papers, citations, or experimental results.
- Prompt-injection defenses separate untrusted retrieved text from executable instructions.
- Agent tool permissions are explicitly allowlisted — no agent has unrestricted database or network access.
- Audit trails log all privileged and consequential academic actions.
- Student code is executed only in sandboxed environments with resource limits.
Open Educational Resources & Licensing
NaySahul University uses only educational materials whose licensing is documented and compatible with intended use. Primary sources include MIT OpenCourseWare, LibreTexts, and original authored content.
For every imported resource we track: source title and URL, author and rights holder, exact license and license URL, attribution requirements, whether adaptation is permitted, redistribution rights, and a retrieval date.
Because future commercialization is possible, the default content collection uses resources compatible with commercial reuse and adaptation. Unclear, incompatible, or mixed-license resources are quarantined for review before any student exposure.
License status tracking
CC BY — permitted (adaptation + commercial reuse with attribution)
CC BY-SA — permitted (share-alike required)
CC BY-NC — restricted (no commercial use)
CC BY-ND — restricted (no adapted redistribution)
No stated license — quarantined for review
Open Source
The entire platform is open-source and developed in public. Contributions are welcome — curriculum proposals, OER metadata, agent implementations, test cases, accessibility improvements, and deployment tooling.
View on GitHub