Graph-based Knowledge Tracing: Modeling Student Proficiency Using Graph Neural Network

Reproducible, production-grade pipelines for modern multimodal vision systems. Efficient VLM adaptation · Embedding-space drift detection · Edge inference · Robustness & safety

LM-Kit Maestro is a secure, innovative desktop application that orchestrates AI agents offline, empowering you to build personalized chatbots with the advanced capabilities of LM-Kit.NET.

Adaptive Model Streaming for real-time video inference on edge devices

[IEEE S&P 22] "LinkTeller: Recovering Private Edges from Graph Neural Networks via Influence Analysis" by Fan Wu, Yunhui Long, Ce Zhang, Bo Li

Enable efficient DNN inference on the edge

LLM chatbot example using OpenVINO with RAG (Retrieval Augmented Generation).

Source code of the paper "Private Collaborative Edge Inference via Over-the-Air Computation".

Lightweight, extensible, and fair multi- DNN manager for heterogeneous embedded devices.

Production Android AI with ExecuTorch 1.0 - Deploy PyTorch models to mobile with NPU acceleration and 50KB footprint

PocketLFM — run Liquid AI's LFM2.5 large language model fully on-device on Android. Offline, private, no cloud. Open-source edge AI via llama.cpp (GGUF). First independent open-source Android app to run LFM2.5 outside Liquid's own Apollo/LEAP.

Research and training stack for AVA — a tool-using, memory-aware virtual assistant targeting 4 GB VRAM. Spans custom transformers, verifier-RL, external memory, multi-domain benchmarks, and Gemma 4 inference optimization.

This project is a wearable navigation aid that combines computer vision, edge inference, and obstacle detection. The system provides audio feedback to assist visually impaired individuals in navigating their surroundings.

Flutter + LiteRT/tflite demo for obj detection

Code for paper "Dynamic Deep Neural Network Inference via Adaptive Channel Skipping"

Arbitrary Numbers

Code for Task-Oriented Communication for Multi-Device Cooperative Edge Inference, IEEE TWC, Jan. 2023.

Analysis, Local Studio, and deployment decision layer for local-first Edge AI inference validation.