Data Science Fellowship Program


Data Science Fellowship Program Group Shot

Launched in 2016, the LSST Discovery Alliance (LSST-DA) Data Science Fellowship Program (DSFP) positions graduate students to meet the scientific challenges of large astronomy datasets.

The DSFP is designed to supplement your graduate education in astronomy by teaching essential skills for dealing with the large data set soon to be produced by the Vera C. Rubin Observatory’s LSST.

Meet our past and present Data Science Fellowship cohorts.

Program Support, Tuition, and Scholarships

Travel for DSFP fellows is covered by generous support from the Brinson Foundation, the Gordon and Betty Moore Foundation, the WoodNext Foundation, and the Research Corporation for Scientific Advancement. The DSFP program tuition is $6,000 for participants; however, full scholarships are available for some students. Selection is not contingent upon having funding for tuition, and all applications are reviewed prior to fellowship offers being made.

Program Overview

The DSFP is a two-year training program designed to teach skills required for Vera C. Rubin Observatory Legacy Survey of Space and Time (LSST) science that are not easily addressed by current astrophysics programs. Fellows learn a wide range of essential skills, including the basics of managing and building code, statistics, machine learning, scalable programming, data management, image processing, visualization, and communication. Our program is a supplement to graduate education, intended to teach students in astronomy-related fields (e.g., astrophysics, cosmology, planetary science, etc.) essential skills for dealing with big data. 

DSFP Class

The DSFP consists of six one-week schools over a two-year period (three per year), each at a different host institution. This gives students ample time to attain skill mastery. As this program is intended as a true supplement to graduate education, fellows make a two-year commitment. They have the opportunity to not only study these topics in far greater depth than traditional schools but also foster new collaborations and professional networks. On top of teaching our students the skills they need for modern survey astronomy, we also aim to create a collaborative, supportive learning environment and work to empower our students to teach the skills they learn to others. 

The Curriculum

Below is a partial list of topics the DSFP covers. Our curriculum is developed and distributed openly. To facilitate the exploration of our materials, we have linked example lessons for the topics below (full material is available via our GitHub repository). We also film every lecture and post the lectures to our YouTube channel.

Software Engineering

  • Building code repositories
  • Object-oriented programming
  • Version control/GitHub
  • Issue tracking
  • Unit tests
  • Continuous integration

Statistics

  • Regression
  • Frequentist vs. Bayesian methods
  • Gaussian processes
  • Generative models
  • Hierarchical models
  • Missing information and selection effects

Machine Learning

  • Unsupervised methods, including density estimation, anomaly detection, feature extraction, and clustering techniques
  • Supervised methods
  • End-to-end automated classification models
  • Deep neural networks

Scalable Programming and Data Management

  • Parallel programming
  • Databases
  • Software profiling
  • Cloud computing

Time Series Analysis 

  • Understanding variable sources with incomplete and noisy sampling
  • Measures of periodicity
  • Gaussian processes
  • The LSST alert stream

Image Processing

  • Noisy astronomical detectors
  • Processing pipelines
  • Position, flux, and shape measurements
  • Hands-on experience with the LSST image-processing software stack

Visualization  

  • Visualization of large dimensional data sets
  • Interactive visualization for exploration
  • Visual hierarchies
  • The effective use of space, color, contrast, and textures

Science Communication 

  • Understanding your audience
  • Effective body language for communication
  • Presentation design principles
  • Using data to tell a story

Program Leadership

The Data Science Fellowship Program is led by:

For questions, use our contact form. Please select Data Science Fellowship Program from the dropdown.

Fellowship Success

By all measures, the DSFP has been successful in its goals:

  • Of the 110 students in Cohorts 1-6 of the DSFP, 99% reported that DSFP contributed to their PhD and/or securing their current position.
  • 80 have received a PhD, 26 are still working toward a PhD, 47 have become postdocs (27 won prize fellowships), and 7 have become tenure-track faculty or permanent staff.
  • The vast majority of past students who received PhDs and did not continue in academia are in data science industry jobs.

LSST Discovery Alliance gratefully acknowledges support for the Data Science Fellowship Program from:

support logos for the Data Science Fellowship Program

https://www.youtube.com/@lsstdadatasciencefellowship144/videos