Nandish Shah

LinkedIn

Nandish Shah

PolyTutor: Multi-Agent Learning for STEM & Language Mastery

PolyTutor is an AI-powered, domain-specific educational platform that uses specialized agent teams to guide learners through subjects like Physics, Chemistry, Math (PCM), and Language Learning. It addresses the lack of personalized, interactive, and multimodal learning experiences in education, especially for complex STEM topics and language acquisition.

Project Goal To demonstrate a modular, multi-agent tutoring system that offers personalized, engaging, and multimodal education for learners at various levels—bridging conceptual clarity, practical problem-solving, and cultural language immersion.

How It Works

User selects a subject (Physics, Chemistry, Math, or Language). System activates relevant agents (e.g., Concept Agent, Visual Agent, Grammar Coach, etc). User engages with interactive modules: Asks questions, solves problems, or explores visual explanations. Core Functionality

Concept Agent: Offers deep explanations and critical thinking support. Problem Agent: Guides users step-by-step through practice problems. Visual Agent: Generates diagrams and visual aids for clearer understanding. Search Tool: Uses vector search over NCERT and other PDF materials. Language Agents: Includes Grammar Coach, Vocabulary Builder, Dialogue Partner, and Culture Expert. Domain Validator Agent: Validates Input belongs to right subject.

FastAPI backend Server Agno Framework (multi-agent orchestration) LanceDB + Embedder (for vector search) youtube_transcript_api & pytube (language learning) DuckDuckGo Search (external fact retrieval) ReasoningTools (agent reasoning support) Custom UI built in React + Tailwind

Yes

No

Yes

Yes