Andrew Chang
- Founder and CEO, American Data Science
- andrew@amdatascience.com
- AI Lead, Golden
- andrew@golden.co
- Mathematical Sciences, Carnegie Mellon University
Hi, my name’s Andrew Chang and I like to build things.
My experience in engineering involves AI/ML, deep learning, natural language processing, and MLOps. I enjoy building out full stack applications on distributed systems with tools like Python, Docker, and Kubernetes.
I am the creator and maintainer of packages and repositories including AdaptNLP, Godel, GoldNLP, ML Docker images, and other tools/services.
Interests. Open Source, MLOps, Natural Language Processing, Accelerated Computing
My experience in engineering involves AI/ML, deep learning, natural language processing, and MLOps. I enjoy building out full stack applications on distributed systems with tools like Python, Docker, and Kubernetes.
I am the creator and maintainer of packages and repositories including AdaptNLP, Godel, GoldNLP, ML Docker images, and other tools/services.
Interests. Open Source, MLOps, Natural Language Processing, Accelerated Computing
Projects
- American Data Science [blog] Simple Jupyter Lab and Hub management platform with pre-configured environment runtimes supporting a variety of tools, libraries, and GPUs.
- Fast Neural Networks (FastNN) [code] A framework for deploying serializable and optimizable neural net models at scale in production via. the NVIDIA Triton Inference Server.
- AdaptNLP [code] [blog] A high level framework and library for running, training, and deploying state-of-the-art Natural Language Processing (NLP) models for end to end tasks. Built atop Zalando Research’s Flair and Hugging Face’s Transformers library, AdaptNLP provides Machine Learning Researchers and Scientists a modular and adaptive approach to a variety of NLP tasks with an Easy API for training, inference, and deploying NLP-based microservices.
- Automated Learning Rate Suggester A learning rate finder for the 1cycle learning policy. Automates the selection of a recommended learning rate in FastAI for each step of training with an interval slide rule technique
- ML Docker Images Docker images for ML/AI applications with containers ranging from CUDA-compatible jupyter servers and environments to Kops/k8s plugins for CI/CD workflows. Docker images for AdaptNLP are also available here
Conferences and Workshops
2020
- State-of-the-art Text Classification Made Easy with AdaptNLP, Open Data Science Conference Europe. [code] [blog] Andrew Chang and Brian Sacash
- State-of-the-art NLP Made Easy with AdaptNLP, Open Data Science Conference East. [code] [blog] Andrew Chang and Brian Sacash