Kin Long Kelvin Lee

Postdoctoral Researcher — Center for Astrophysics

Harvard & Smithsonian

https://laserkelvin.github.io

kinlee@cfa.harvard.edu • +1 857-505-9734 • 161 Kelton Street, Allston MA 02134

---

Professional Experience

Since Feb 2017

Center for Astrophysics | Harvard & Smithsonian—Cambridge, MA

Postdoctoral Research Fellow

• Developed machine learning models and open-source frameworks for automated spectroscopic analysis and molecule discovery in space and in the laboratory; workflow reduced analysis time from months to weeks.
• Co-authored 16 publications; six as lead author, one as project leader, and remainder as part of large teams.
• Lead successful grant proposals as co-investigator from the Smithsonian Institution, National Science Foundation, and NASA; raised over $500,000 USD in public funding.
• Mentored early career researchers on scientific and numerical Python, and reproducible workflows.
• Reviewed new research articles in major publications: role as referee/reviewer on 13 manuscripts.

Aug 2016‒Feb 2017

University of New South Wales—Sydney, Australia

Postdoctoral Research Fellow

• Researched photochemistry of atmospheric molecules; destruction of pollutants under UV irradiation by lasers.
• Lead the experiments, analysis, and dissemination of multiple research projects through peer-reviewed publications and oral presentations at international conferences.
• Published five publications in major peer-reviewed journals; three as lead author.
• Mentored undergraduate students on various theoretical and experimental research projects.
• Developed open-source tools for automated analysis of ion images and trajectory simulations.
• High accuracy quantum chemical calculations of photolytic reactions of pollutants; quasi-classical trajectory simulations of how molecules dissociate.

Education

2020

Coursera

Deep Learning Specialization

Five deep learning courses taught by Andrew Ng, deeplearning.ai; completed April 3, 2020.

2013‒2016

University of New South Wales—Sydney, Australia

Doctor of Philosophy in Chemistry

Title: Spectroscopy and Photodissociation of Small Atmospheric Molecules under the supervision of Professor Scott Kable and Professor Meredith Jordan.

2008‒2012

University of Sydney—Sydney, Australia

Bachelor of Science; First Class Honours in Chemistry & Plant Sciences

Title: Roaming Reaction Dynamics in Small Aldehydes under the supervision of Professor Scott Kable and Professor Meredith Jordan.

Selected Open-source Contributions

Python

PySpecTools is a library I developed to help analyze broadband spectral data with an emphasis on reproducibility and collaboration.

RotConML is a project that uses quantum chemical datasets to teach probabilistic deep learning models to identify molecules from spectroscopic data.

Learning Neural Networks is a resource I developed for teaching beginners deep learning from scratch.

SpectraViewer is a lightweight Dash web app for inspecting and performing signal processing on legacy laboratory data.

My Github repository contains all of the coding projects I have worked on.

Selected Recent Publications

• Identification of unknown molecules using probabilistic deep learning models
  • Developed high throughput, probabilistic neural network architectures to identify unknown molecules with rotational spectroscopy from computational chemistry data.
• Accuracy and uncertainty benchmarking of quantum chemical methods with Bayesian methods.
  • Determined systematic uncertainties with low-cost electronic structure theory using Hamiltonian Monte Carlo models.
• Developed open-source tools for analyzing broadband spectral data
  • Developer of PySpectools, a Python library that helps manage analysis of rotational spectra consisting of hundreds of spectral features and distinct species.
• Conservation of zero-point energy in quasiclassical trajectory simulations
  • Created a new, low-cost method for “patching” zero-point energy leakage in molecular dynamics simulations, applied to roaming reaction dynamics.

Skills & Expertise

Languages

English and Cantonese as native languages.

Fluent in conversational Japanese.

Computing

Distributed computing workflows on national and institiutional HPC platforms.

Massively parallel quantum chemistry calculations using CFOUR, Gaussian, Psi4.

Source control managment with git and dvc.

Reproducible environments with conda and docker.

General object-oriented programming and development with Python 3.

Linux shell workflows (bash, zsh).

Data analysis

Exploratory data analysis and data pipeline design with scientific Python and Julia stack.

Machine/Deep learning models with PyTorch, Keras, scikit-learn.

Probabilstic model development with pymc3, pyro, and Flux.jl.

Written communication

Author of 21 peer-reviewed articles for expert audiences; 98 citations to date. [Link to Google Scholar]

Review Editor of open-access journal “Frontiers of Astronomy and Space Science”.

Writer on Medium and TowardsDataScience for general audiences.

Oral Presentations

Presenter at SciPy 2020.

Presented scientific results at over 18 international conferences in Chemistry and Astronomy.

Presented workshops on reproducible Python and code practices to undergraduates at the Center for Astrophysics

Harvard & Smithsonian.

---

kinlee@cfa.harvard.edu • +1 857-505-9734 • 161 Kelton Street, Allston MA 02134