Skip to main content

About NOBOGYAN

A multi-agent education platform built for the 15th China Software Cup

Mission

NOBOGYAN is a multi-agent education platform that replaces the traditional "one-size-fits-all" curriculum with a data-driven approach. Every resource, learning path, and tutoring interaction is generated based on individual student profiles.

Built for the 15th China Software Cup (iFlytek Track), NOBOGYAN demonstrates how multi-agent AI systems can transform higher education through intelligent orchestration and continuous adaptation.

The Story

Traditional learning platforms deliver identical content to every student, regardless of their prior knowledge, learning style, or goals. This leads to:

  • Boredom for advanced learners covering known material
  • Frustration for struggling learners skipping foundational concepts
  • Inefficiency with static curricula that can't adapt

NOBOGYAN was born from a simple question: What if we could build a system where 15+ AI agents collaborate in real-time to create a unique learning experience for every student?

The Agent Swarm

Content Generation Agents

The Creators

Scholar

Lecture notes and reading materials

Mapper

Interactive mind maps and knowledge graphs

Sage

Adaptive quizzes and assessments

Director

Video scripts with narration

Architect

Programming exercises and code challenges

System Intelligence Agents

The Coordinators

Orchestrator — Task delegation and coordination
Tutor Engine — Real-time Q&A and explanations
Path Planner — A* algorithm for optimal learning paths
Recommender — Content-based and collaborative filtering
Faithfulness Checker — Hallucination detection
Gate Agent — Milestone validation and unlocking
Evaluator — Performance analysis and insights
Profile Extractor — Learning profile inference
Gap Detector — Knowledge gap identification
Vision LLM — Image and diagram analysis
Voice Agents — ASR and TTS processing

Core Innovation: The Orchestrator Pattern

Unlike single-prompt AI systems, NOBOGYAN uses a coordinating agent that:

  1. Analyzes the learner's profile and current context
  2. Decomposes requests into sub-tasks
  3. Dispatches specialized agents in parallel
  4. Aggregates and validates outputs
  5. Delivers a cohesive, multi-format response

This pattern eliminates reliance on massive prompts and enables true multi-modal generation.

Key Features

FeatureDescriptionInnovation
Conversational ProfilingExtracts 6-dimension learner profiles through natural chatNo forms or explicit questionnaires
Multi-Agent Generation5 agents generate notes, mind maps, quizzes, video, code simultaneouslyParallel async with faithfulness checks
Adaptive Path PlanningA* search over knowledge graphs with real-time replanningDynamic based on performance signals
Multimodal TutoringText, voice, image, and diagram support in one interfaceStreaming SSE with context management
Assessment & AnalyticsLLM-based evaluation with automatic remediation triggersClosed-loop feedback to profiling

Performance Targets

Time to First Token

<800ms

Design target

Resource Generation (Text)

<5s

Design target

Resource Generation (Video)

<30s

Design target

Profile Update Latency

<2s

Design target

Path Replanning

<3s

Design target

Concurrent Users

50+

Design target

Team

VELOX Studio

A team building multi-agent educational technology for the 15th China Software Cup (iFlytek Track).

Founded

2026

Focus

AI-powered adaptive learning

Our Vision

We believe in democratizing quality education through AI. NOBOGYAN is designed to serve:

  • •Educational institutions seeking adaptive learning platforms
  • •Students who need personalized learning experiences
  • •Professionals looking to upskill efficiently
  • •Organizations investing in employee development

"An entire AI team, learning with you."