Chris Miles
Data scientist and mathematical physicist
University of Michigan
Ph.D. Physics 2018
M.S. Applied Mathematics 2015
Massachusetts Institute of Technology
B.S. Physics 2010
Programming: Python, Jupyter notebook, Sklearn, Numpy, Pandas, Flask, Pytorch, Keras, AWS, SQL, MongoDB, LaTeX
Subject expertise: Numerical optimization, Numerical differential equations, Time-series analysis, Reinforcement learning, Stochastic processes
San Francisco
Data Science Fellow
Jan. 2019 to Current

Developed 5 data projects as part of a 3-month immersive program.

University of Michigan
Ann Arbor
Graduate Student Researcher
June 2012 to May 2018

  • Developed  fluid simulations and optimization models in Python, and performed numerous other computational and analytical studies for the optimization of fluid mixing.
  • Used a collection of optimal control methods to discover that diffusion can limit the mixing effectiveness of incompressible flows in some cases.

On-Ramp Wireless: Communications Physical Layer
San Diego, CA
Systems Engineering Intern
Summer 2011 to Fall 2011

  • Used Python to investigate signal processing data to determine the presence of signal interference between ORW's wireless network and WiFi networks.
  • Developed a decision tree classifier to help avoid signal interference.

Continental Tires R&D: Pattern, Contour, and Layout
Hanover, Germany
Mechanical Engineering Intern
Fall 2010 to Winter 2011

  • Used MATLAB to develop simulations of the likelihood of tire wear and damage.
  • Contributed to early concept-phase development of tire tread pattern designs for upcoming products.

Data Projects
Converting Python to C++ code using LSTMs

Used an LSTM sequence-to-sequence architecture to convert Python code to C++ code.

Detecting respiratory symptoms in breathing audio recordings

Used a classic KNNs for the detection of respiratory symptoms such as wheezing and crackling in the recordings of breathing. 

Time-series analysis for stock prediction

Explored classical statistical models such as ARIMA for time series analysis to predict long-term (months) and short-term (1-day) price movements.  A simple autoregressive model motivated by physical spring (harmonic oscillator) dynamics is explored. 

Deep Q-Network training of a food-seeking reinforcement learning agent

Implemented a Deep Q-Network algorithm to train an agent that seeks food and avoids poison in Unity's 3D virtual environment. 

Repository of multi-agent reinforcement learning environments

Created an open source python package with multi-agent reinforcement

learning environments. The package includes classic 2-player matrix

games and multi-player Snake environment.