Justin Lê, Ph.D.

Learning-Based Modeling of Weather and Climate Events Related to El Niño Phenomenon via Differentiable Programming and Empirical Decompositions

Information-theoretical statically verifiable data privacy guarantees in machine learning algorithms and deployments.

Backend engineer for a large-scale Haskell web-facing application simulating entire corporate internal and external networks for the purpose of realistic mock cyber attacks. Direct team responsibilities involved bridging user-facing interfaces with complex scheduling and execution of applications and processes (namely, attack agents) across multiple virtual machines, analyzing and assessing results of attacks in a data visualization pipeline, and coordinating with cloud services and APIs. In the capacity of a Haskell programmer, implemented large type-safety and type-directed code generation initiatives company wide for greater internal correctness guarantees and more robust channels of communication between front-end and back-end.

Developed an ensemble-based Machine Learning system for forecasting and predicting frequency and intensity of power outages for major Energy and Utility company providing for over 3 million people. Developed mathematical models based on stochastic principles for analysis and pre-processing of data. Worked with Neural Network, Self-Organizing Map, Stochastic models, and ARIMA models to provide a ensemble forecast. Worked also on developing an on-line platform to manage updating models and generating predictions as weather data was submitted.

Involved in the development of the technology, underlying mathematics, and user interface for MathDB, an abstracted data store used for real-time streaming data analysis. Assisted in the promotion and integration of MathDB technology in different capacities. Directed the planning of educational programs in Machine Learning and Data Science aimed for university students and industry professionals in Ukraine.

Modeling complex topologies of superconducting quantum interference devices for magnetoscopic applications, and implementing efficient, parallel numerical simulations under those models for calibration and experimentation.

Authored and maintained open-source backprop, backprop-learn platform/library for the Haskell language, providing automatic differentiation in support of differentiable programming and machine learning based projects. Currently used by many in the Haskell open source community to build richer data science platforms. Additionally, authored the opto platform for efficient extensible numerical optimization.

Applied principles of the Feynman Path Integral Formulation of Quantum Mechanics to create real-time high-performance, parallelizable numeric simulations in multiple languages, including C++ and Fortran, for live analysis and exploration of ground state quantum systems.

Maintaining a top Functional Programming and Haskell blog with 50,000 pageviews per year, appearing multiple times on the front page of high-visibility platforms such as Hacker News. Topics include mathematical models, functional programming, and dependently-typed and type-safe programming.

Atmospheric Research https://www.sciencedirect.com/science/article/pii/S0169809517300157 (January 2017)

https://blog.jle.im/entries/series/+functional-models.html (May 2018)

Compose Conf 2019, New York, New York http://talks.jle.im/composeconf-2019/ (May 2019)

AOGS 13^th Annual Meeting, Beijing, China http://talks.jle.im/aogs-2016/ (August 2016)

Study of statistical time series analysis and statitiscal models for studying and analyzing time series data with mathematical rigor. Applied to financial time series analysis data, comparing the efficacy of different statistical models.

The computational study of various mathematical models and simulation techniques in historical financial data, specializing in comparative market analysis and currency exchange.

Survey of remote sensing techniques, including the acquisition, aggregation, processing, analysis, and physical considerations of geophysical satellite data. In-depth look at a wide range phenomenology including meteorological anomalies, dust, fire, and anthropological impacts.

Study of the modern state of high performance computing and big data. In-depth look at parallel and concurrent computing through various approaches, architectures, and network topologies. Applying cluster and grid computing algorithms to compute-intensive tasks.