SATPC0036254 Tab 04 4 SOW.pdf
PDF 187 KB Posted
- Solicitation number
- 80NSSC24879852Q
About this file
This document is a Statement of Work (SOW) for the development of a Harmonized Landsat Sentinel-2 (HLS) Geospatial AI Foundation Model (GeoFM) to advance earth observation science. The key objectives are for the contractor to provide research and science expertise to support the development of the AI foundation model, including enhancing the model architecture, developing strategies for model improvements, creating labeled training data, documenting benchmarks, and supporting fine-tuning and performance validation. The contractor will also provide outreach and communication support, such as organizing workshops and training, to educate the geospatial science community on the GeoFM capabilities.
The related federal contract opportunity is a sole source requirement from the National Aeronautics and Space Administration Shared Services Center to develop and test the HLS GeoFM for science-driven advances in earth observation. The period of performance is from September 1, 2024 through February 28, 2026, and the place of performance is at facilities at Arizona State University.
View the file
Other files
| File | Type | Posted |
|---|---|---|
| SATPC0036254 Tab 07 Capability Statement SAM.gov.pdf | ||
| SATPC0036254 Tab 06 RDSS..pdf |
On GovTribe
Work with this file
- Download the original file
- Contacts named in this file
- Similar government files
- Ask GovTribe AI about this file
Text version
Statement of Work for
Harmonized Landsat Sentinel-2 (HLS) Geospatial AI Foundation Model (FM) for Science-driven Advances for Earth Observation
Background and Objective:
Artificial intelligence (AI) model development is playing an important role in advancing our understanding of science. Foundation models (FM), in particular, provide flexibility for developers to pretrain a single model and then customize it for specific downstream tasks.
The NASA Inter-agency Implementation and Advanced Concepts Team (IMPACT) at the Marshall Space Flight Center (MSFC) is developing a foundation model for HLS Geospatial AI Foundation Model (commonly referred to as GeoFM) and is seeking support in designing, building and evaluating the model to advance and explore a wide range of tasks within the field of geospatial science (i.e., global climate change, polar environment and its sustainable development, terrain morphology, and urbanization, etc.).
Requirements: The contractor shall:
Provide research and science expertise to support the development of a FM in HLS Geospatial applications.
a. Assist the AI team within IMPACT to develop enhancements to the model’s architecture.
b. Develop strategies for potential enhancements to the GeoFM architecture, in areas such as multi-scale learning, location-embedding learning, multi-modal learning, and better incorporation of temporal dependency.
c. Develop strategies such as spatial cross-validation and categorizing testing data relevant to different geographic regions, landscapes, and climate zones to evaluate the GeoFM’s transferability and generalizability in geospatial scientific applications.
d. Develop high-impact functional use cases to test, evaluate, and utilize the GeoFM.
e. Create and provide access to labeled, science training for each downstream use case.
f. Create or provide documentation for new or existing benchmarks to evaluate the performance of the FM.
g. Support the development of fine-tuning approaches for the downstream use cases.
h. Assist the AI team within IMPACT to validate and evaluate the performance of the FM on downstream use cases.
Provide outreach and communication support aimed at educating and training the science community on the capabilities of the GeoFM.
a. Organize and host workshops, tutorials, webinars and/or hackathons with the geospatial science community.
b. Train scientists and developers on the capabilities of the AI FM related to the developed downstream use cases.
c. Assist the IMPACT team with disseminating research findings through technical reports, conference presentations, and journal publications.
d. Assist with publicizing the AI FM results to the scientific community as an open-source science collaboration.
Development Cycle: The contractor shall:
Follow the FM development iteration cycles shown below.
Deliverables: The contractor shall deliver:
a. Access to labeled, science training data for each downstream use case.
b. Documentation for new or existing benchmarks to evaluate the performance of the FM.
c. Training & tutorial materials to be used during workshops, webinars and/or hackathons with the atmospheric science community.
d. Research findings as technical reports, conference presentations, and journal publications.
Place of Performance:
Facilities at the Arizona State University
Period of Performance:
September 1, 2024 through February 28, 2026.
File details come from the government source that posted it. Updated .