About brandon
Hi! I am a Machine Learning Research Leader at the University of California, Riverside. I have spent the past 7+ years mastering machine learning concepts and applying them to real-world engineering projects. I am passionate about statistics, machine learning, and data science, especially when making a tangible difference in the world. This is what drove me to get a Ph.D. in Electrical Engineering and Computer Science at the University of California, Riverside. Over the past seven years, I have used machine learning methods to solve problems across the United States Power Grid. I am proud to have worked on various topics, including Theft Detection, Anomaly Detection, Topology Classification, and Generating of Power System Data from Noise. In solving these problems, my team and I have used a heavy amount of machine learning and data analysis - usually relying on very large datasets. I have a solid knowledge of statistics and statistical machine learning theory, having invented and proven new theorems in Information-Theoretic Machine Learning Theory and Large Deviations Theory in statistics. I have mastered the art of building large-scale deep neural networks in both PyTorch and Tensorflow. My ability to produce and test more shallow learning methods via Scikit-Learn, Pandas and NumPy are also quite strong. I would say that I have a keen intuition for knowing when shallow, interpretable learning is more appropriate than deep. Most recently, I have led a research team consisting of myself and six graduate students through a two-year machine learning project solving major power system problems tasked by the U.S. Department of Energy. This research project has been quite successful, as we beat the performance of every other team involved! I am excited at the opportunity to expand my breadth of knowledge by working on challenging, novel problems. The beauty of machine learning is that, as the practitioner increases their breadth, their depth becomes stronger!
Key Skills
Experience
Machine Learning Engineer
CurrentHabitat Energy
Postdoctoral Machine Learning Researcher
University Of California, Riverside
Led a research team solving major bulk US power grid problems using real, large-scale Synchrophasor data. Specifically, Created a realistic set of synthetic Synchrophasor events via Tensor Decomposition and Generative Adversarial Networks. Inception score - 95%. Holds under expert scrutiny. Created a classification model for Power System Event Signatures via a deep convolutional residual neural network with Mutual Information Loading. F1 score - 97%. Created an automatic Power System Event Signature Mining / Label Generation System via clustering. Classification models trained on these labels yield highly confident predictions (Shannon Entropy <0.2). Created an Online method for Synchrophasor Missing Value Replacement via Tensor Decomposition and manifold-projected gradient descent. Errors kept under 10% during highly volatile event periods. Created a method of detecting the start of a power system event via a Bidirectional Generative Adversarial Network trained during normal periods. F1 score - 95%.
Graduate Student Researcher
University Of California, Riverside
Solved several problems in distribution systems troubling power utility companies by utilizing real smart meter data. Specifically, Created a classification model for Phase Identification in Power Distribution Systems via information Loading. Classification Accuracy - 97% while training on just 5% of the data. Developed a method to detect Electricity Theft from Smart Meters via linear regression of power data from the voltage data of neighbors. Increases detection rate by a factor of 20 without introducing false positives. Performed detailed financial analyses of Battery Storage Valuation via their participation in Wholesale Power Markets. Performed Battery Storage Degradation Modelling and Optimization in Power Markets under Degradation. In conjunction with the above, I also contributed several mathematical proofs to the field of Information-Theoretic Machine Learning Theory. Specifically, Created Tighter Generalization Error Bounds in Statistical Machine Learning Theory. Proved the existence of bounds that are linear in the information stored by a neural representation rather than exponential. Related the rate of information loss accrued by data sampling to well-known data quality metrics.
Education
University Of California, Riverside
Doctor Of Philosophy
Ucla
Bachelor Of Science
Interested in connecting with brandon?
Sign up for NinjaHire to send a connection request.
Common Questions
What is brandon's expertise?
brandon specializes in Machine Learning Engineer, with expertise in data science, deep learning, keras, machine learning, python.
Where is brandon located?
brandon is based in moreno valley, california, united states.
How much experience does brandon have?
brandon has 14+ years of professional experience.
How can I contact brandon?
You can connect with brandon through NinjaHire by signing up for a free account.