I'm a research scientist at the Lawrence Berkeley National Laboratory (LBNL), working between the Energy Technolgies Area (ETA) and the National Energy Research Supercomputing Center (NERSC) on physics-informed and generative machine learning models. Previously I have been a postdoctoral scholar with the Civil and Environmental Engineering Department at MIT, and held a research scientist affiliation with the Grid, Mobility, and System Integration group with the Applied Program at SLAC. My work is on computational methods (with a focus on machine learning/AI) for analyzing and optimizing energy, infrastructure, industrial, and natural resources systems. Of particular interest are methods to integrate, process, and harness novel satellite, sensor, and human-generated data.
Previously, I have spent two years as a senior data scientist at C3 IoT, a machine learning startup in the San Francisco Bay Area, working on failure prediction and anomaly detection algorithms for large-scale infrastructure systems and industrial machinery. I obtained my PhD in Electrical Engineering at Stanford University working on machine learning methods for smart grid applications.
- Machine learning, computer vision, remote-sensing, operations research
- Energy, natural resources, and cities.
Albert, A., Maasoumy, M. Predictive segmentation of energy consumers. Applied Energy, 2016. pdf
Albert, A., Rajagopal, R. Finding the right users for thermal demand-response: an experimental evaluation. IEEE Transactions on the Smart Grid, 2016. pdf
Albert, A., Rajagopal, R. Thermal profiling of residential energy consumption. IEEE Transactions on Power Systems PP(99), 2014. pdf
Albert, A., Rajagopal, R. Cost-of-service segmentation of energy consumers. IEEE Transactions on Power Systems, 6(29), 2014.pdf
Albert, A., Rajagopal, R. Smart meter driven segmentation: what your consumption says about you. IEEE Transactions on Power Systems, 4(28), 2013. pdf
Armel, C., Gupta, A., Shrimali, G., Albert, A. Disaggregation: the Holy Grail of Energy Efficiency? Energy Policy, (2012). pdf
Vogt, J., Albert, A., and Marghitu, O.: Analysis of three-spacecraft data using planar reciprocal vectors: methodological framework and spatial gradient estimation, Ann. Geophys., 27, 3249-3273, 2009. pdf
peer-reviewed conference papers
Albert, A., Kaur, J., Gonzalez, M. C. Using convolutional networks and satellite imagery to identify patterns in urban environments at a large scale. ACM KDD, Halifax, Nova Scotia, 2017. arXiv
Albert, A., Rajagopal, R. Strategic scheduling of energy consumers. IEEE Conference on Decision and Control, December 2015, Osaka, Japan. pdf
Albert, A., Gebru, T., Ku, J., Kwac, J., Leskovec, J., Rajagopal, R. Drivers of variability in energy consumption. ECML-PKDD DARE Workshop on Energy Analytics. pdf
Albert, A., Rajagopal, R. Building dynamic thermal profiles of energy consumption for individuals and neighborhoods. IEEE Big Data Conference, 2013. pdf
Albert, A., Schafer, A. Demand for freight transportation in the U.S.: a high-level, data-driven view. Transportation Research Forum, Annapolis, MD, March 2013. pdf
Gomes, N., Merugu, D., O'Brien, G., Mandayam, C., Yue, T., Atikoglu, B., Albert, A., Fukumoto, N., Liu, H., Prabhakar, B., Wischik, D. “Steptacular: an incentive mechanism for promoting wellness”. NetHealth, 2012. pdf
Albert, A., Rajagopal, R., Sevlian, R. Power Demand Distributions: Segmenting Consumers using Smart Meter Data. ACM SenSys (BuildSys) 2011. pdf
thesisProblems, models, and algorithms for data-driven energy demand management. Stanford University, 2014. pdf
Technical advisor and founding team member at Myriad Sensors
Myriad Sensors builds Pocket Lab, a sensing and analytics platform for performing on-the-go scientific experiments focused on physical sciences. Our first product targets students, educators, and citizen scientists.
MSc and BSc theses
Multi-point data analysis for magnetospheric multi-spacecraft missions (MSc Thesis)
For my MSc Thesis research with Prof. Joachim Vogt at Jacobs University Bremen I developed a novel method for processing time series of three-point magnetospheric satellite data. Part of this work has been published (another part sadly never made it to the editor…) slides
Efficient calculations of cosmic particle fluxes in paleomagnetospheres (BSc Thesis)
My BSc thesis with Prof. Joachim Vogt investigated a computationally-efficient method for calculating charged particle fluxes in magnetic fields by considering trajectories of particle populations that were relevant to final flux integration.
conference presentations on MSc and BSc research
Albert, A., Vogt, J., Marghitu, O. Three-spacecraft gradient calculations: accuracy studies and vorticty calculations. International Association of Geomagnetism and Aeronomy (IAGA) 2009 General Assembly, Sopron, Hungary (oral presentation). pdf
Albert, A., Vogt, J., Marghitu, O. Analysis of three-spacecraft data using planar reciprocal vectors. American Geophysical Union Fall Meeting 2008, San Francisco (poster presentation). pdf
Albert, A., Vogt, J., Zieger, B. Efficient calculation of cosmic particle fluxes in paleomagnetospheres. German Geophysical Society Meeting, Bremen 2005 (oral presentation).
Selected internship research projects
Inferring high level personal mobility choices from sparse gps data
In summer 2010 I worked at IBM Almaden Research Center on developing an integrated system that uses GPS information provided by smartphones to analyze and model individuals’ mobility choices. The system collects GPS data from Android and iPhone devices via native apps, sends it automatically to a server, performs analytics, and lets the user view and interact with the results in a web portal. We were interested in extracting high-level travel patterns and dominant choices that could be used as input for a map model-aware route prediction tool.
Lighting authoring effort analysis
We analyzed the authoring effort for several Pixar movies (Ratatouille, Wall-E, and UP) for lighting versus animation effects. Insights of the statistics were used to support research decisions in front of Pixar management (results and written materials are confidential).
Multi-sensor electric field change networks for lighting research
At NASA MSFC I worked in the atmospheric and lightning physics group on developing a software package for analyzing electric field change data from an array of sensors in the Huntsville, AL area. poster
Numerical investigation of the kelvin-helmholtz instability
As a summer REU project at the University of California, Santa Barbara, we proposed a novel regularizing scheme for the Birkhoff-Rott equation of an ideal vortex sheet, based on an exponential smoothing term, and performed numerical experiments of the sheet's stability. slides
Efficient simulation of many colliding deformable objects
I participated in RIPS (Research in Industrial Projects for Students) at UCLA, where I worked in the Pixar team on developing a computationally-efficient approach for simulating many colliding deformable objects for Pixar's Ratatouille. pdf
Selected internship technical reports
Albert, A., Bitzer, P., Christian, H., Blakeslee, R. BLITZER: A data analysis system for lighting research. NASA Marshall Space Flight Center Intern Colloquium periodical 2008. pdf
Albert, A., Atanasova, D., Creech, K., Scott, J. Desingularization of a vortex sheet using a variable smoothing parameter. University of California, Santa Barbara Mathematics Colloquium, 2007. pdf
Albert, A., Bachmayr, M., Wiener, J., Effinger-Dean, l. Efficient simulation of many colliding deformable objects for set-dressing and animation. Technical report for Pixar Animation Studios, 2006. pdf
Selected class projects
Visualizing the U.S. freight network
This is a small dashboard I put together for Jeff Heer's visualization class. It describes inter- and intra-state flows of freight and some of its determining factors. Note that for some browsers the zoom setting might need to be adjusted.