Machine Learning Systems


Alphan Altinok
Brian Bue
Michael Burl
Selina Chu
Gary Doran
Emily Dunkel
Tara Estlin
Raymond Francis
Robert Granat
Anantha Ravi Kiran
Jack Lightholder
Steven (You) Lu
Lukas Mandrake
Yuliya Marchetti
Umaa Rebbapragada
Peyman Tavallali
Kiri Wagstaff




The Machine Learning and Instrument Autonomy (MLIA) Group creates software solutions to hard problems requiring data mining, knowledge discovery, pattern recognition, and automated classification and clustering. The underlying emphasis is on building systems based on learning algorithms. We conduct basic research as well as develop applications leading to one-of-a-kind proof of concept systems. Our focus is on the automated analysis of scientific data generated by NASA and JPL instruments, on the development of technologies for adaptive systems, and on enabling technologies for autonomous spacecraft.

  • Now hiring: Come help us create machine learning systems for science and spacecraft!
    • Machine Learning Research Data Scientist (Ph.D.):
      Contact with your CV and a statement of interest
  • COSMIC Zooniverse
  • Citizen science with COSMIC (Content-based Onboard Summarization to Monitor Infrequent Change): We are building an ambitious autonomous system to find and track changing events on a planetary scale around Mars. You can help by studying fascinating Mars images and labeling points of interest through this easy-to-use interface: COSMIC Zooniverse.

  • Umaa Rebbapragada was co-author on two papers related to the first detection of gravitation waves at LIGO. She is one of over 1500 co-authors on "Localization and broadband follow-up of the gravitational wave transient GW150914" by B. P. Abbott et al. of the LIGO collaboration. She is also a co-author on a paper released by the intermediate Palomar Transient Factory (iPTF) titled "iPTF Search for an Optical Counterpart to Gravitational Wave Trigger." Both papers document how iPTF was involved in the follow-up effort to analyze regions of the sky containing GW150914. iPTF did not find any follow up candidates associated with the gravitational wave trigger. However, the papers document the end-to-end discovery process in following up a gravitational wave detection. JPL-developed software for transient vetting enables the discovery of supernovae in iPTF imagery. The JPL team that created the "real-bogus" transient vetting software consists of Umaa Rebbapragada, Brian Bue, and Gary Doran.

  • The intermediate Palomar Transient Factory (iPTF) made a first significant discovery of a young astronomical transient (currently hypothesized to be a very early detection of a Type Ia supernova) via the real-bogus transient vetting that was deployed to the image subtraction pipeline hosted by the the Infrared Processing and Analysis Center (IPAC) for iPTF. This discovery was reported in an Astronomer's Telegram titled "iPTF Discovery of a Young Transient in a Tidal Tail of NGC 5221." Umaa Rebbapragada, Brian Bue, and Gary Doran are listed as co-authors to acknowledge work on the JPL-developed "real-bogus" transient vetting software developed for the IPAC pipeline. This discovery is worthy of note because the original image subtraction pipeline (in operation at the National Energy Research Scientific Computing Center) did not find this very young supernova candidate.

    Data Ordering Genetic Optimization (DOGO)
  • Lukas Mandrake received the Runner-Up award for the 2015 NASA Software of the Year competition for the Data Optimization via Genetic Ordering (DOGO) System, which automatically ranks data by its estimated quality and has been adopted by the OCO-2 mission. You can learn more about DOGO by viewing the NASA SOY DOGO presentation and the NASA Tech Brief on DOGO.
data analysis image labeling

rock finding active learning