Work closely with software engineering professionals to get ready for Rubin LSST data
Contents:
I. Call for proposals
II. Incubator Program
III. Planning a proposal that fits with LINCC Frameworks
IV. Two-stage proposal process
The LINCC Frameworks team at LSSTC, Carnegie Mellon University (CMU), and University of Washington (UW) invites scientists to participate in the LINCC Frameworks Incubator Program. The goal of the Incubator Program is to engage the scientific community in the creation of software tools for the analysis of Rubin LSST data. To that end, we will support teams of researchers to test and help expand early-stage analysis software being developed as a part of LINCC Frameworks using their own exciting scientific investigations with existing surveys or simulated data. Selected Incubator teams will be invited to send a Team Lead to CMU or UW for an extended visit of up to three months. In addition to involving the research community in the creation of valuable analysis tools for survey science, we expect the LINCC Frameworks Incubators to lead to cutting-edge scientific publications and to position Incubator participants for success in their future funding and observing proposals.
At this time, we invite proposals for the first Incubator session, to be held in mid-2023. Please consider attending weekly virtual 'office hous' on Thursdays at 12 pm EST with a member of the LINCC Frameworks team in the wekks leading up to the proposal deadline.
Those in the LSSTC Slack workspace can use the public #lincc-incubator-help channel for questions about the proposal process; members of the LINCC Frameworks leadership are monitoring this channel.
Key Dates:
- Virtual Information Session: 2023 January 26. You can find the recording here.
- Stage-one proposal deadline: 2023 February 21
- Stage-two conversations with selected proposal teams: 2023 March
- Expected announcement of first Incubator teams: 2023 April
Funding, computational resources, and collaboration with software engineers. The LINCC Frameworks Incubator Program, which starts in mid-2023, will provide dedicated funding and collaborations with software engineers to proposal teams who pitch compelling projects to be carried out in collaboration with the LINCC Frameworks team. A successful Incubator project will tackle a computational challenge for eventual LSST data analysis by developing or testing open-source LINCC Frameworks software that can be applied to existing simulations and/or precursor data. Incubator project teams will be provided with approximately $20k for research support (e.g. travel, graduate student salaries, faculty summer salaries) to enable the project team's collaboration with LINCC Frameworks software engineer(s) and research scientist(s). Computational resources will be available to support Incubator projects.
Funding Details. LSSTC is able to provide funds to both US-based and international PIs. We do not, however, request budgets along with stage-one Incubator proposals. We will request budgets at a later stage from the PIs of successful proposals. The funding cap will be $22,500 per Incubator team, and there will be tight restrictions on the portion of these funds that can be used for local administrative costs.
First Incubator session. The first 2023 Incubator program will run for three months between June through September (exact dates are flexible to accommodate participants’ availability). Two project teams will be selected for this session – one to collaborate primarily with the LINCC Frameworks personnel based at CMU and one to collaborate primarily with the LINCC Frameworks personnel based at UW. We anticipate implementing three Incubator sessions per year throughout 2023-2026, with two new project teams selected for each session (i.e., six new Incubator projects selected each year). We plan to update the program throughout this time as we receive feedback and to release regular, open calls for proposals.
Incubator project team composition and size. Each project must include a Project Lead who will work with LINCC Frameworks software engineers and/or data scientists for the equivalent of approximately 16 hours per week, including attending a weekly meeting with all the project teams. The Project Lead can be a student or postdoc, and can be a different person than the Principal Investigator (PI, who will receive and oversee the grant). Although we understand that some Incubator projects might be carried out within a larger group, such as an LSST Science Collaboration or an in-kind team contributing to the Rubin SIT-COM effort, we consider the Incubator team to consist of only the (likely) small number of people who will be actively involved in the Incubator project.
We anticipate that we will support both remote and in-person participation in the Incubator Program. On the proposal form, proposers are asked their preference for working fully remote, hybrid, or in-person. Preference will not impact project selection.
Collaborative approach. Collaboration in shared virtual or physical spaces is important for deep technical engagement and provides opportunities for cross-pollination among multiple concurrent projects. Thus, Incubator projects are not “for-hire” software jobs — the Project Lead will work in collaboration with LINCC Frameworks software engineers and research scientists. Each Project Lead will be responsible for successfully moving the project forward, with the Frameworks personnel using software industry standard approaches to provide guidance on methods, technologies, and best practices as well as general software engineering during the three-month incubator period. As in other academic collaborations, the contributions of the LINCC Frameworks personnel and the role of the Incubator program should be properly attributed in resulting talks, publications, software releases, etc.
Incubators are meant to kick off new scientific analyses and approaches, all within an open-source coding environment to benefit broad sectors of the LSST research community. The three-month timeline provides a concentrated window for both ramp up and collaboration, and the scientific research will likely extend a few months beyond the end of the Incubators. The scope of proposed projects should be commensurate with this three-month timeline, with an eye toward the full research question.
Recommended programming experience. To maximize the value of the three-month period of close collaboration, the Project Lead should have proficiency in reading and writing code. We recommend at least one year of experience programming in Python. Depending on the focus of the Incubator project, the technical work may range from minor modifications of existing LINCC Frameworks code (e.g., adding a new analysis function) through the development of entirely novel algorithms.
Expected outcomes or deliverables. We will ask the following from each of the selected Incubator teams:
- A few-paragraph write-up at the start of their project to be posted on the LINCC Frameworks website.
- Open-source code on GitHub, as per standard practice throughout LINCC Frameworks
- A notebook showing their analysis
- An update to their write-up at the end of the project
- A talk about their project from a member of their team, for example at that team member’s home institution
III. Planning a proposal that fits with LINCC Frameworks
LINCC Frameworks focus areas. Successful proposals will focus on problems that involve computationally intensive analysis and that drive novel scientific research. For the first year, Incubator projects should be closely aligned with the primary LINCC Frameworks focus areas:
- Frameworks for querying and analyzing catalog data (e.g. crossmatching catalogs and correlation functions)
- Robust analysis of time series data at LSST scale
- Generalizing photo-z algorithms and frameworks
- Detection and characterization of minor planets (e.g. light curve fitting, and detection of TNOs with precursor data)
More information about LINCC Frameworks personnel, along with their expertise and interests can be found here, while a description of ongoing activities in the LINCC Frameworks team may be found here.
Planning your Incubator project. One way to define a project would be to think of the science you are excited to do with early LSST data, then think of starter projects that would be a first step in the direction of your LSST project using existing precursor or simulated data. In other words, you should be able to do your Incubator project now, given the opportunity to develop early/prototype software for the science use case. Below we present a few examples of potential projects. These are not meant to be comprehensive, but rather seed ideas for proposals.
- Implement a novel, computationally-intensive spatial analysis (e.g. whole survey N-point correlations, statistical analysis of source matching across different catalogs, proper motion computation) using the LINCC Frameworks LSDB package. Project participants would implement new metric code needed for the analysis and conduct the analysis using existing data sets.
- Integrate LSDB with the Rubin Science Platform (RSP) and identify optimizations needed for efficient cross-matching of catalogs with very large numbers of sources to enable tests to be carried out during Rubin commissioning, with the LINCC Frameworks team providing expertise on optimization. Then use the optimized code for a project that involves cross-matching of multiwavelength precursor survey datasets to test photo-z estimation, deblending, or other image- or catalog-level processing in the two datasets.
- Implement and test a specific science-driven time series feature computation or algorithm on precursor data using the LINCC Frameworks lsstseries package. Project participants would implement the new metric or algorithms code within lsstseries and conduct a novel study on a large-sized precursor survey data.
- Implement and optimize a new, computationally intensive photo-z quality assessment metric motivated by the needs for AGN or galaxy science in the context of DESC’s RAIL/qp software packages. Then use precursor or simulated data to conduct a comprehensive comparison of how current photo-z codes perform using that metric versus using existing cosmology-motivated metrics.
- Adapt the KBMOD software package to search for trans-neptunian objects in other large-scale precursor surveys. Project participants would write input adapters to convert data from the existing surveys to LSST’s format and then run the search.
- Develop a computationally efficient technique for identifying activity within images of minor planets. Project participants would write the image analysis code or train models (if applicable), and conduct an analysis of the effectiveness and scalability on precursor surveys.
IV. Two-stage proposal process
Stage-one proposals for the mid-2023 Incubator session must be submitted by 2023 February 21, via the web portal link. The first Incubator teams will be announced within approximately two months.
Eligibility. The program is open to anyone whose LSST-specific research can be significantly advanced by intensive collaboration with a software engineering expert. For the first Incubator session, however, we will be particularly open to proposals that seek to test existing software (whether developed by the proposers, by the LINCC Frameworks team, or other groups) in the current LINCC Frameworks focus areas. Moreover, we expect that active LSST Science Collaboration members will be able to leverage their experience to develop compelling Incubator proposals. Over the full life time of the LINCC Frameworks Incubator Program, more than three-years, we aim to select a scientifically balanced portfolio of Incubator projects, and for at least half of the selected Incubator projects to be led by a researcher or student from an LSSTC member institution.
Stage-one proposals. To submit a stage-one proposal, please answer the questions on the Incubator proposal web portal and upload a short project description of one to two pages describing the scientific goals of your project, the relevant precursor and/or simulated datasets you plan to use, and the expected technical challenges associated with your project.
Stage-two conversations. In March, LINCC Frameworks personnel will provide constructive feedback to all proposal teams and initiate more extensive follow-up conversations with a subset of proposal teams whose projects the selection committee deems to be particularly strong and appropriate for the first Incubator session. A primary goal of these stage-two conversations will be to more thoroughly assess the feasibility of potential projects. We expect to announce the selection of the first Incubator teams in 2023 April.
Selection Criteria. A successful Incubator proposal will clearly identify both the needed tools and the scientific questions to be answered. It will explain how the technical component of the project is critical to delivering LSST science. It will lead to a publishable scientific result in the near term. In more detail, review of Incubator proposals will be based on the following criteria:
- Desire to engage with LSST. Participants are available and engaged or intending to engage in preparation for LSST, as demonstrated through, e.g., working at an LSSTC member institution or active participation in an LSST Science Collaboration.
- Project definition. Project is well-defined, both technically and scientifically. It is likely to lead to measurable outcomes, such as a quantifiable increase in some capability of LINCC Frameworks tools or an improvement in accuracy of some output, etc.
- Feasibility. Project is feasible to carry out now given existing datasets and computational resources. Proposal demonstrates that the main, closely collaborative part of the project can reasonably be accomplished in three months. It describes a plan for completing the scientific analysis and publication of a paper within a reasonable time frame after the three-month Incubator.
- Compatibility with LINCC Frameworks. For at least the first Incubator session, the proposed project is well aligned with LINCC Frameworks focus areas and/or the capabilities and interests of the Frameworks scientific staff (see https://www.lsstcorporation.org/lincc/frameworks for focus areas and current activities in those areas). A LINCC Frameworks Incubator would be uniquely valuable for catalyzing project success – e.g., there should be a clear indication that something about the project is computationally non-trivial.
- Community Impact. Project benefits a broad and diverse research community. The proposal team describes an inclusive plan to communicate about their work and the tools to the community.
Some good proposals will not meet all criteria, and some great proposals may not be selected due to limited bandwidth or expertise of our team members. With multiple rounds of incubators and at least two calls for Incubator proposals expected each year, there will be opportunities to try again!
