🚀 I built Saidrix AI Tutor — an AI-powered live learning platform where 10+ specialized AI agents work together to create a personalized learning experience.

Instead of building another chatbot that simply answers questions, I wanted to build a system that can understand a student, identify knowledge gaps, build a personalized course, teach live, review projects, and remember progress over time.

🧠 What’s running behind Saidrix AI Tutor:

→ Intake Agent — Understands what the student wants to learn and builds a learner profile through intelligent, topic-specific questions.

→ Knowledge Profiler — Runs adaptive diagnostic rounds and evaluates actual answers to determine what the student really knows.

→ Course Maker + Project Planner — Creates personalized courses, chapters, lessons, and practice projects grounded in a RAG knowledge base.

→ Lecture Maker — Plans and generates complete lessons, including diagrams. Generated visuals go through geometry checks and vision-model review before reaching the student.

→ Project Reviewer — Pulls student projects from GitHub or ZIP files, reviews files in parallel, checks requirements, and returns line-specific feedback with a quality score.

→ Long-Term Memory System — Distills each learning session into verified student context so future lessons can continue from where the student left off — without inventing information about the learner.

🎙️ And one of my favorite parts: the Live Voice Tutor.

Built with Python + LiveKit, it can teach lessons out loud, handle student interruptions naturally, answer questions, and then continue teaching from where it stopped.

⚙️ The multi-agent system is orchestrated using LangGraph.js + TypeScript and designed to be LLM provider-agnostic, allowing models from OpenAI, Claude, or Gemini to be switched through configuration.

Building this taught me an important lesson about production AI:

The hard part isn't making an LLM generate something. The hard part is building a reliable system around it.

Structured outputs, schema validation, RAG grounding, LLM call gating, rate limiting, bounded repair loops, and fail-safe agents became just as important as the models themselves.

This is what I've been building at Saidrix — moving toward AI that doesn't just answer students, but actually teaches them. 🎓🤖

🌐 aitutor.saidrix.com

#AI #AIAgents #EdTech #LangGraph #RAG #TypeScript #Python #LiveKit #LLM #ArtificialIntelligence #SoftwareEngineering #Saidrix