CSUN Senior Design • 2025–2026

Smarter Recycling, Powered by AI

RecyKool is a multimodal, agentic AI system designed to reduce downtime in Waste Management’s conveyor sorting system. The platform combines an AI chatbot interface on Apple Vision Pro, an NVIDIA Omniverse Isaac Sim digital twin of the sorting line, material characterization to identify items that may jam equipment, and gesture recognition to detect unsafe worker movements. Together, these components provide real-time operational insight that helps improve line reliability, minimize stoppages, and support safer working conditions.

VLM Understanding

RAG/KG Retrieval

Twin Simulation

Material Characterization / VLM Developer

GitHub • LinkedIn

Project Overview

What RecyKool Does

RecyKool helps users make correct disposal decisions using multimodal AI. A user can submit text, images, or combined inputs to identify materials, understand common contamination mistakes, and receive grounded guidance that explains recommended actions. The system is designed to be practical, transparent, and aligned with real-world operational constraints.

Major Features: material identification, disposal guidance, retrieval-grounded explanations, structured knowledge reasoning, and context awareness via digital-twin concepts.

🧠

Technologies

FastAPI, Docker, Python

👓
Platform

                                Vision-based UI concepts and multimodal interaction for a modern guidance experience.
                            

♻️
AI Approach

                                VLM + RAG + Knowledge Graph reasoning to reduce hallucination and improve reliability.
                            

Jump to Demo

Skills Demonstrated

Technologies and capabilities developed during the project

Applied multimodal AI techniques including vision-language models, retrieval-based reasoning, and structured knowledge representations to generate grounded explanations.

AI / Vision-Language Models

Python, Jupyter, ML workflows, RAG, knowledge graphs

Developed and interacted with a real-time Omniverse digital twin of a conveyor belt system to simulate operations and visualize system behavior.

Digital Twin Systems

NVIDIA Omniverse, simulation workflows, system modeling

Built and tested an immersive frontend experience for Apple Vision Pro, focusing on spatial UI interaction and usability.

Frontend Development

VisionOS, Swift, spatial UI design, testing

Demo & Contact

Screenshots, video, and project links

🎓

CSUN Senior Design

California State University, Northridge
Course Projects Page: Back to Course Page

🔗

Project Links

Repo: GitHub
Demo: YouTube
Documentation: Project Notes

🧩

Tech Stack

API: FastAPI • Python
Infra: Docker • Cloud Storage (if used)
AI: Vision-Language • RAG • Knowledge Graph