I am a research engineer with over 2 years of experience working with deep learning methods and models. I am currently lead research engineer for a joint research project on applied deep learning in cosmology.
I enjoy working on challenging problems and working with others who are as enthusiastic about the developments and research of AI as I am.
I am looking for research-engineer or software-engineer type roles within AI/ML, applied or research, and would love the opportunity to learn more about your platform and learning tasks and discuss how I can benefit your team!
- Git / Source Control
- IPython Notebook
- Machine learning
- Scikit Learn
- shell scripting
August 2014 May 2018
Bachelors: computer science at University of British Columbia
August 2009 May 2012
Bachelors: East-Asian Literature and Cultures at University of California, Berkeley
August 2016 August 2017
Research co-op student at Honda Research Institute, Japan
As a research engineer at Honda Research Institute, I spent a year working closely with a research scientist on cutting-edge unsupervised visual learning tasks. Working with advanced deep learning technology, my only resources for improving our models was academic literature, and implementing and experimenting with new techniques and methods from these papers became a daily responsibility.
Rarely supported by existing library assets available in deep learning frameworks, and I regularly hand-crafted complex functions from current SOTA literature at the lowest level, including their backprop gradients in python and kernelwise CUDA C++ so they could be utilized by the model. I worked with Chainer, Keras-theano, TensorFlow, NumPy, SciPy, CUDA, Jupyter notebooks, and matplotlib on linux, and distributed model training over 4 TITAN Xs.
Research Engineer Lead at Academic Research
I am lead research engineer for a continuing collaboration between researchers at UBC, Berkeley, and CMU for applying deep learning in cosmology. There are two cosmologists, two machine learning experts (the project lead scientist, and myself), and one particle physics scientist. The project’s goal is to accelerate N-Body simulations using machine learning to model the evolution of the universe under the highly non-linear influence of dark matter.
We use a non-volumetric (or set-based) dataset which has atypical constraints that has led to the original development of some novel shift-invariant and permutation equivariant network layers. We use TensorFlow, Chainer, PyTorch, scikit-learn, NumPy, SciPy, matplotlib, and CUDA.