Senior Design GenAI Subsystem
Built the generative AI subsystem for a multi-platform cognitive coaching application. Designed and implemented RAG system, artifact generation, and interactive chat with conversation context.
System Architecture
GenAI subsystem for a multi-platform cognitive coaching application combining RAG, artifact generation, and interactive chat. Built with FastAPI backend, Qdrant vector storage, and multi-provider AI integration.
Documents → Vector DB → RAG → Artifact Generation → Chat Service → Multi-Platform APIThree-layer context system combines RAG retrieval, conversation summaries, and recent message history for coherent conversations.
Key Features
RAG System
Retrieval-augmented generation system with Qdrant vector storage, document ingestion pipeline, and context-aware retrieval for educational content.
Artifact Generation
Built artifact generation system creating flashcards, multiple-choice questions, and insights with schema validation and template-based generation.
Interactive Chat Service
Developed chat service with three-layer context system combining RAG retrieval, conversation summaries, and recent message history.
Multi-Provider Integration
Integrated multiple AI providers (Ollama for local, Purdue GenAI API for cloud) with automatic provider selection and fallback mechanisms.
Technical Details
FastAPI Backend
Built FastAPI backend with RESTful endpoints for artifact generation, chat interactions, and document management with error handling.
CLI Interface
Created Typer-based CLI with commands for artifact generation, chat interactions, document ingestion, and system configuration with tab completion support.
Vector Storage
Implemented Qdrant vector database integration with persistent collections, semantic search, and document retrieval for RAG operations.
Impact & Results
Delivered GenAI subsystem enabling educational artifact generation and contextual AI assistance for cognitive coaching platform
Key Achievements
Architected and implemented RAG system with Qdrant vector storage and document ingestion pipeline
Built artifact generation system (flashcards, MCQ, insights) with schema validation and template system
Developed interactive chat service with three-layer context (RAG + Summary + Recent messages)
Created FastAPI backend with CLI interface for artifact generation and chat
Integrated multiple AI providers (Ollama, Purdue GenAI API) with automatic provider selection
Technical Highlights
- • GenAI subsystem architecture and implementation
- • RAG system with multi-layer context management
- • Artifact generation with schema validation
- • Multi-provider AI integration with automatic selection
- • CLI and API interfaces
- • Full-stack AI system design