Junming Huang

Versatile AI researcher and engineer transitioning from Cybersecurity to frontier AI. Currently a Research Member at the DUAL Research Group (University of Sydney), exploring Distributed AI (Federated Learning, edge deployment), LLM Evolution (RL, fine-tuning/inference), and Neural Dynamics (Neural ODEs, Flow Matching). Co-authored papers at CAIT 2025, with work under review at IEEE Internet of Things Journal and IEEE Networking Letters.

My research interests include resource-efficient AI, parameter-efficient fine-tuning (PEFT) in federated learning, and CV-based conservation technology — including a joint project with BirdLife Australia on individual identification of critically endangered Regent Honeyeaters using Computer Vision.

I also maintain a self-driven open-source LLM project on Hugging Face: J1mm7/mini_lm, re-implementing the full MiniMind pipeline from pre-training through SFT, LoRA, and RLHF alignment.