Jennifer Pollack


Contact Information
E-mail: jennifer.pollack [at] cea [dot] fr
Office: 281
Affiliation: IRFU/DAp-AIM

About me

I'm a former cosmologist turned software engineer and consultant, aspiring to learn industry's best practices in the areas of software development and project management to help large-scale experiments successfully achieve its scientific objectives.  

For the past 11 years, I've been working on various aspects of the European Space Agency's Euclid Mission, one of the next generation cosmology surveys aiming to shed light into the mysterious nature of dark energy. The experimental design of Euclid is one-of-a-kind with two payload instruments onboard to carry out simultaneous photometric and spectrophotometric observations, respectively, at the visible and near-infrared wavelengths of the electromagnetic spectrum for its two core science probes: weak gravitational lensing and galaxy clustering. Throughout my career with Euclid, I've provided support in terms of software development for both science experiments as well as towards improving Euclid's internal communication infrastructure.

My first contributions to the mission began in the design of a scalable software algorithm for Euclid's Science Ground Segment to characterise the clustering properties of galaxies in the non-linear regime using a complex class of high order statistics called the bispectrum.  Measurements of the galaxy bispectrum contain a wealth of information about the nonlinear growth of structures in the dark universe driven by the gravitational collapse of dark matter structures at small scales and halted by cosmic expansion at large scales due to dark energy.  These statistics help us peer into the dark universe via the light-emitting objects we can see: galaxies, by constraining galaxy bias which is how the distribution of galaxies in the universe reflects the clustering pattern of the total matter density field dominated by dark matter.  

I then transitioned to another part of the Science Ground Segment pipeline to work in the exciting domain of pixel-level image processing working on simulations of the upcoming Vera C. Rubin Observatory Legacy Survey of Space and Time which will feature the largest CCD camera ever built with ~3 billion pixels to map  ~18,000 sq. degrees of the sky out to large imaging depths. Joint processing of Euclid and Rubin photometry will help provide precise measurements of the redshifts of galaxies for Euclid's gravitational weak lensing program (described further below).  I also worked jointly on the integration of the Rubin Observatory pixel processing software into the Euclid Science Ground Segment pipeline.  I helped guide Euclid leads to better strategies in dealing with the massive dataset anticipated for Rubin coupled with the host of complexities in image quality arising from the instrumental and atmospheric effects to be addressed by Rubin's elaborate, yet state-of-the-art photometric calibration campaign.  

And here, at CosmoStat I've ventured into yet another stage of the Science Ground Segment pipeline working on modelling the instrumental response of Euclid's VIS camera commonly referred to as the point spread function (PSF), which characterises the amount of blurring and distortion introduced by the telescope optical system.  Euclid's VIS camera has been especially designed to enable extremely precise measurements of billions of galaxy shapes for Euclid's weak gravitational lensing experiment: a 1.5 billion galaxy survey campaign to measure the tiny distortions induced by the gravitational effects of large-scale structures in the Universe on distant background galaxies.  The information in the weak gravitational lensing signal measured at different redshifts can be used to reconstruct the expansion history of the universe in order to learn more about cosmic acceleration.  And hence, modelling accurately Euclid's point spread function to disentangle its effect from the weak lensing signal is a pivotal step to Euclid achieving its core science aims.  

Over the past few years, CosmoStat has pioneered a new approach to PSF modelling with the development of WaveDiff (Liaudat et al., 2023), a differentiable data-driven wavefront-based PSF modelling framework in TensorFlow.  It's a hybrid methodology that constructs a model of the PSF across the telescope's field-of-view by combining a forward parametric model of the wavefront with non-parametric data-driven techniques used to correct forward model errors in the wavefront space.  WaveDiff has been especially designed for the Euclid experiment to ensure the mission can achieve and exceed its science performance requirements for weak lensing. My contributions to the project over the past year has been in refactoring the software to higher quality standards in software design, testing, and maintainability, as precursor steps needed to integrate WaveDiff into the Euclid Science Ground Segment.  In the next milestones, I will be working towards adding new features like modelling the dichroic beamsplitter which induces spurious wavelength dependences into the PSF model that varies spatially and will also begin carrying out initial performance tests with early Euclid data.

Finally, in the past year, I rediscovered one of my earlier aspirations in the area of communications (a field I once was pursuing, endeavouring to become a sports journalist). A good communication network is essential for large collaborations with many interfacing teams to thrive, allowing groups to working efficiently, meet deadlines and milestones, and reduce resource costs.  It also fosters a collaborative and supportive environment for members at all career stages.  It's good for science.  I've been assisting in the Euclid Consortium's Internal Communication and Management group as a consultant and as lead and developer for the Euclid Consortium Slack workspace with >1500 members.  As lead, I've written the EC Slack Management and User Guidelines and the EC Slack workflow plan for administrating the EC Slack workspace.  As developer, I've built software applications for performing user identity management for the EC Slack workspace as well as Bot applications to help with automation and information sourcing.  In the past annual Euclid meeting in Copenhagen 2023, my colleague Lucie Baumont and I hosted Euclid's first Sprint Day, featuring nine events throughout the day where teams held brainstorming & hack sessions, and software demos.  Over 70 members of the Euclid Consortium participated in this first-ever event.