INTELLEA — Multi-Agent Adaptive Learning PlatformINTELLEA — Multi-Agent Adaptive Learning Platform

Adaptive learning needs to track what a student actually knows and respond to it, which means agent workflows that hold state across a session, stay grounded in real course material, and fail without taking the rest of the platform down.

Architected the platform as four independently designed modules, with modular agentic pipelines on a plugin-based architecture and production-oriented patterns: stateful orchestration, retrieval grounding, fault isolation. Personally built ATLAS, the platform intelligence unit — shared knowledge base, assignment-creation engine, and virtual tutoring component — combining Bayesian Knowledge Tracing with LangGraph-orchestrated agent workflows and Graph RAG over a Pinecone vector store. Backed the automated grading and document-ingestion pipelines with Apache Kafka for an event-driven architecture.

Designed and integrated the entire CI/CD, building Jenkins pipelines and owning full AWS deployment. Integrated OpenAI APIs, ElevenLabs TTS, and Firebase, with fallback handling for graceful degradation.