About Me
Hello, I'm Lai Wei, an Engineering Manager at AWS AI. I'm passionate about Artificial General Intelligence (AGI) and have extensive experience in distributed systems, large-scale model training, machine learning frameworks and systems.
My inspiration comes from my favorite show, "Person of Interest" (first produced in 2011), where the Machine, an advanced AI, was trained by its creator, Harold, using Reinforcement Learning with Human Feedback(RLHF). This show sparked my interest in the development of AGI and motivated me to contribute to the field.
Throughout my career, I have had the opportunity to work on cutting-edge projects and collaborate with talented teams that share my vision for a future where AGI plays a critical role in improving our world.
Resume
Check out my resume for more information about my professional experience and skills.
What's New
- CyberBunny: GPT4o controlled robot car
- Resolvr: GenAI for DevOps, open source project
- Contributed Claude 3 backend for Self Operating Computer, demo
- CFO Copilot: AI copilot for enterprise CFOs, open source project
- Submission to OpenAI's Preparedness Challenge: Proof of Concept AI Misuse
Blogs & Talks
Here are some of my past blogs, talks, and works I contributed to:
- Tip and tricks building custom GPT
- Build High Performance ML models using PyTorch 2.0 on AWS
- LG AI Research Develops Foundation Model Using Amazon SageMaker
- Amazon SageMaker 大规模机器学习实践
- AWS and NVIDIA achieve the fastest training times for Mask R-CNN and T5-3B
- How Netflix uses Deep Java Library (DJL) for distributed deep learning inference in real-time
- Detecting pneumonia from chest x-ray images in Java
- Using DJL to Deploy a Pre-Trained Deep Learning Model from Apache Kafka®
- MXNet 1.5.0 release
- Apache MXNet (incubating) adds support for Keras 2
- Train using Keras-MXNet and inference using MXNet Scala API
- MXNet开放支持Keras，高效实现CNN与RNN的分布式训练
- Introduction to Deep Learning for Developers (slides)
- RECSYS 2017: Transfer Learning for Personalized Content and Ad Recommendation