Alex Chen

Santa Clara, CA

I am Alex (Wei) Chen, a Director at Qualcomm. Before joining Qualcomm through its acquisition of Nexa AI, I founded Nexa and served as its Founder, CEO, and Chief Scientist, with the goal of making generative AI more efficient and accessible on everyday hardware.

At Nexa, our team built the NexaSDK — an open-source toolkit for running AI models on edge devices that has been widely adopted by the developer community. I was the main technical author behind much of this work, including the Octopus model series, OmniVLM, OmniAudio, and NexaQuant. Along the way, we worked closely with engineering teams at NVIDIA, AMD, Microsoft, Google, IBM, and Intel, and deployed with enterprise partners like Geely, HP, Lenovo, and İşbank.

My interests sit at the intersection of AI systems, LLM/VLM inference, hardware-software co-design, and robotics. I’m especially drawn to the question of how we build intelligence that can act in the physical world. I hold a PhD from Stanford University (2024) and a Bachelor’s from Tongji University (2019).

Selected Work

NexaSDK

Runtime

A GenAI inference runtime that runs models on NPU — 8,000+ GitHub stars and #1 on GitHub Trending.

Hyperlink

Application

A local AI agent that lives inside your computer — it searches across all your files, fully private. 2.1M views and 30K users in two months, featured by NVIDIA, Microsoft, Qualcomm, and AMD.

Octopus

Model

The world's first AI agent model built to run directly on hardware — function calling across PCs, phones, wearables, automotive, and robots — part of a model line that trended top-3 on Hugging Face, featured by Google I/O and the Google DeepMind blog.

For industry recognition of this work — from silicon vendors, model labs, and platforms — see the recognition page.

This blog is where I think out loud about AI, robotics, and the systems behind them. If something sparks a question or a conversation, I’d genuinely love to hear from you.

news

Mar 13, 2026	Joined Qualcomm as Director following the acquisition of Nexa AI.

latest posts

May 11, 2026	Understanding Kernel Design from a Mathematical Perspective
Apr 13, 2026	Gated Delta Networks: Improving Mamba2 with the Delta Rule
Apr 05, 2026	Universal Manipulation Interface: In-The-Wild Robot Teaching Without In-The-Wild Robots

selected publications

arXiv

AutoNeural: Co-Designing Vision-Language Models for NPU Inference

Wei Chen, Liangmin Wu, Yunhai Hu, and 8 more authors

arXiv preprint arXiv:2512.02924, 2025

arXiv Bib HTML

@article{chen2025autoneural,
  title = {{AutoNeural}: Co-Designing Vision-Language Models for {NPU} Inference},
  author = {Chen, Wei and Wu, Liangmin and Hu, Yunhai and Li, Zhiyuan and Cheng, Zhiyuan and Qian, Yicheng and Zhu, Lingyue and Hu, Zhipeng and Liang, Luoyi and Tang, Qiang and others},
  journal = {arXiv preprint arXiv:2512.02924},
  year = {2025},
}

arXiv

Octopus v2: On-device language model for super agent

Wei Chen and Zhiyuan Li

arXiv preprint arXiv:2404.01744, 2024

arXiv Bib HTML

@article{chen2024octopusv2,
  title = {Octopus v2: On-device language model for super agent},
  author = {Chen, Wei and Li, Zhiyuan},
  journal = {arXiv preprint arXiv:2404.01744},
  year = {2024},
}