Prototypes

Early-stage prototypes used to validate system ideas across collaborative inference and conversational AI infrastructure. (Some demos are maintained on student project pages.)

Real-time Multi-device Collaborative Inference (Nvidia Jetson)

Collaborative Inference

This prototype demonstrates pipeline-style inference split across multiple edge devices. It is a lightweight implementation intended to validate cross-device execution and dataflow for LLM inference.

ModelMeta Llama-3.2-3B-Instruct (FP32)
Hardware3× NVIDIA Jetson Orin Nano
FabricTP-Link 5-Port Gigabit switch
SoftwareHuggingFace (re-engineered) + custom PyTorch inference engine

Note: current prototype focuses on multi-device execution; it does not yet model network variability or bandwidth-aware scheduling.

Conversational AI Infrastructure Prototype

AI Infrastructure

An internal prototype demonstrating an end-to-end AI application for conversational behavioral health support.

Our proposal was selected as one of 3 awards out of 59 submissions in a highly competitive internal review.

FocusEnd-to-end AI application supporting behavioral health coaching
StatusInternal demo (not publicly released)

This system is currently maintained as an internal research prototype for behavioral health.