NAIRR Grant Proposal for Smart Aquaculture/Intelligent Fisheries (SA/IF) in the MOC with AI and Edge Observability

The Green Reef Foundation (GRF), led by Dr. Benjamin Branch, aims to solve the problem of AI workforce stimulation of a digital twin concept as a vital workforce skill gap cited by Honeywell (2025). This effort will build earth science relations that will strengthen such outcomes applicable to national interests, such as, the US Air Force pipeline needs as described by Booz-Allen (2025).

We are using a vibe coded open-source Smart Aquaculture/Intelligent Fisheries (SA/IF) smaq.computate.org digital twin platform digital twin platform using Red Hat OpenShift AI for earth science and ecological forecast literacy outcomes. The project's focus on Dynamic Data Driven Applications Systems of digital twins and AI aligns with the dynamic data sciences needs of earth science data, as discussed by Yu.Zhen Lin, Quinoaxi Qi, et al. (2025) and Malik, S., Rouf, R., Mazur, K., & Kontsos, A. (2020). The integration of event-driven architecture, as stated by Red Hat (2019), further enhances its capabilities.

Our proposed work will utilize NAIRR resources around the following use cases:

1. Understanding fish population and ocean health.
2. Tracking fishing boat movements.
3. Monitoring fish species landed and their weight.
4. Tracking processing and delivery times.
5. Providing net-zero waste solutions for seafood.
6. Promoting SA/IF innovation through community outreach, startup support, and IoT applications.

Such a tool could build capacity for Freshwater & Estuarine Aquaculture Pilot, Fisheries Innovation, AI-Driven Environmental Research, Workforce Development & Education and Open Science & Data Sovereignty and more globally competitive US Aquaculture and Fisheries outcomes. Additional activities may include (e.g., fish kill prediction, PFAS detection, shellfish health monitoring) and experimentation with Louisiana (marine), Delaware (estuarine/freshwater), Blue Institute & Blue Innovation Labs in Boston (urban coastal), and the Ecological Forecasting Initiative community as research nodes where research, internships, and Red Hat OpenShift AI training can emerge.

Plan and timeline

Our existing open-source SA/IF digital twin platform, with a developer guide and workbench, was presented at the Vibe code, a data-driven website with generative AI for SA/IF, at the Red Hat OpenShift AI workshop in May 2025. This allows immediate deployment in OpenShift AI within the MOC.

The Green Reef Foundation introduced its SA/IF research and development in March 2025 at the US North East Decisions Institute (NEDSI) conference, showcasing its SA/IF platform built with AI vibe-coding technology from computate.org, which accelerates development by 33 times.

Estimate of cloud resources

This Red Hat OpenShift project allocation will support 5 to 10 individuals from the Green Reef Foundation, universities, or other developers collaborating on the SA/IF digital twin. For 5 workbenches, estimated resources are:

• 26 CPUs
• 72 Gi RAM
• 400 Gi storage
• 2 NVIDIA A100 GPUs GPUs for AI Observability Metrics Summarizer.

NAIRR Resources Required for SA/IF

The SA/IF digital twin promotes data sharing and accuracy, contributing to NAIRR Pilot focus areas:

• NAIRR Secure: The SA/IF platform uses zero-trust fine-grained access control with Keycloak, Hashicorp Vault, and External Secrets Operator for privacy and security.
• NAIRR Software: The SA/IF platform is a complex software system with services across OpenShift nodes, utilizing Zookeeper Cluster Manager, a distributed PostgreSQL database, and Apache Solr. Scalable SA/IF APIs and dashboards collect IoT edge device metrics.
• NAIRR Classroom: The Green Reef Foundation aims to train employees of SA/IF facilities, librarians, data scientists, high school clubs, teachers, small supply chain businesses, and military personnel in open-source AI engagement.

We will integrate the open-source AI Observability Metrics Summarizer, requiring an NVIDIA A100 GPU in the MOC. The Summarizer needs authentication/authorization for its UI, which could be addressed by the AI Telemetry and SA/IF platform's zero-trust access control.

Support needs

The SA/IF project is developing a FiWare data space for ocean, fish population, map, water quality, and ecological forecasting data. Our goal is to use AI to demonstrate the value of a new aquaculture/fisheries economy based on a digital twin. We intend to use the AI Observability Metrics Summarizer to query IoT edge device metrics and secure the chatbot with fine-grained access control. Using Red Hat resources (RamaLama, Inference Server) and IBM Granite Instruct and geospatial models, we aim to enhance understanding and leadership in seafood industry best practices.

We seek assistance from IBM, MOC, and AI Alliance in selecting an appropriate AI model for multilingual dialogue applications and chatbots, initially proposing IBM granite-3-2b-instruct or granite-3-8b-instruct. In addition, we want to explore other IBM Granite Instruct, Instruct Lab, geospatial, and time series models from IBM and the AI Alliance that could benefit SA/IF or ecological forecasting.

Team and team preparedness

The open-source vibe coding process for the SA/IF platform uses Smart Data Models from the FIWARE organization, allowing easy forking and extension for sovereign data solutions in various domains, locations, and organizations. This will be a valuable educational tool for associate degree programs, self-paced community programs, or Air Force advanced placement in digital twin training.

The project research question is: What is the most effective way to introduce digital twin technology with AI workflows to build earth science outcomes? Such should promote investigators’ sharing of information and ideas; coordinate ongoing or planned research activities; foster synthesis and new collaborations; develop community standards; and in other ways advance science and education through communication and sharing of ideas. Moreover, this approach includes workshops, online training, quarterly reports, and GRF publication activity with participants. Our community of researchers includes the Ecological Forecasting Initiative (EFI) and its Coastal and Marine Ecological Forecasting (CMEF) working group, Blue Institute & Blue Innovation Labs (BI&BI), Delaware State University (DSU), its Water Quality Lab, and all GRF network institutions such the InterAmerican University of Puerto Rico, Delaware State University, Mississippi State University, Carteret Community College, and Southern University.

Our four-step approach:

• Step 1: The Schedule and Planning Committee will plan the following: The program design consists of 4 quarterly workshops or hackathons with coordinated projects. Workflows will be split into Aquaculture, Fisheries, and location sustainable workflows.

  This effort will work to support a workshop at the EFI 2026 Conference in Toronto.

  • Workshop 1 qtr: AI foundations, Research questions development, Granite Models training, other training needs
  • Workshop 2nd qtr: IoT metrics, Fleet control at Delaware State University, Water Quality Lab and EFI/CMEF AI training roadmap
  • Hackathon 3rd qtr: Community Outreach and demonstration as recruiting tool practice
  • Workshop 4th qtr: BI&BI supported projects by this GRF AI community
• Step 2: Literacy outreach to: the Ecological Forecasting community, Blue Institute & Blue Innovation Labs (based in Boston, MA), and Green Reef Foundation network institutions, explaining free compute resources. Newsletters, conference announcements, and GRF network invites will recruit participants, benefiting AI innovation in Aquaculture, Fisheries, awareness for net-zero waste, shelf stable and specialized crops. Promotion will be sent out 35 days prior to program launch and co-developed with our partners.
• Step 3: Data workflow application development: Participants will identify an Aquaculture, Fisheries, or location-specific workflow that adds value to earth science relations, collaboration, advanced training, and research reproducibility via open source, open data, open science, and open source infrastructure. This will generate 15-25 earth science real-world problems or AI workflows for recruiting and training.
• Step 4: Showcase and marketing of AI in earth science workflows: Benefit the earth science community in SA/IF or earth science workflow through showcasing and testing workflows before presentation or making virtual presentation demos. A data provenance sheet for workflows will be the target outcome.

References

1. The live SA/IF platform: https://smaq.computate.org/
2. The SA/IF developer getting started guide: https://smaq.computate.org/en-us/view/article/deploying-smart-aquaculture-on-openshift-local
3. The SA/IF workbench automation: https://github.com/computate-org/smart-aquaculture-workbench
4. The SA/IF Code Generation workshop to the North East Decisions Institute (NEDSI) conference in March 2025 in Hershey, Pennsylvania: https://smaq.computate.org/en-us/shop/event/smart-aquaculture-nedsi-2025
5. The Vibe code a data-driven website with generative AI for SA/IF in Red Hat OpenShift AI hands-on workshop at the Red Hat Summit May 2025 in Boston, Massachusetts USA: https://smaq.computate.org/en-us/shop/event/smart-aquaculture-red-hat-summit-2025
6. AI vibe-coding open source technology provided by computate.org: https://computate.org
7. The open source AI Observability Metrics Summarizer: https://github.com/rh-ai-kickstart/openshift-ai-observability-summarizer?tab=readme-ov-file
8. The AI Telemetry platform: https://aitelemetry.apps.obs.nerc.mghpcc.org/
9. NEON ecological forecasting data: https://data.neonscience.org/data-products/explore
10. Ecological Forecasting Initiative — Coastal and Marine Ecological Forecasting: https://ecoforecast.org/coastal-and-marine-ecological-forecasting/
11. Unleashing multimodal magic with RamaLama: https://developers.redhat.com/articles/2025/06/20/unleashing-multimodal-magic-ramalama#multimodal
12. Earlier article published in Red Hat Research Quarterly magazine May 2024: Moving ecological forecasting from supercomputer to cloud: why and how https://research.redhat.com/blog/article/moving-ecological-forecasting-from-supercomputer-to-cloud-why-and-how/
13. Yu.Zhen Lin, Quinoaxi Qi, et al, (2025, August) A Dynamic Data Driven Applications Systems (DDDAS)-Based Digital Twin, IEEE https://arxiv.org/pdf/2501.00051
14. Malik, S., Rouf, R., Mazur, K., & Kontsos, A. (2020, October). A dynamic data driven applications systems (DDDAS)-based digital twin IoT framework. In International Conference on Dynamic Data Driven Application Systems (pp. 29-36). Cham: Springer International Publishing
15. Booz-Allen. “Building DOD’s Largest-Ever Digital Twin of Its Kind.” Building DOD’s Largest-Ever Digital Twin of Its Kind, Booz Allen Hamilton, 17 July 2025, https://www.boozallen.com/insights/digital-twin/building-dods-largest-ever-digital-twin-of-its-kind.html
16. Honeywell. “Advanced Digital Twin Tech Helps Close Today’s Industrial Skills Gap.” Energy & Sustainability Solutions, 24 Sept. 2025, https://pmt.honeywell.com/us/en/about-pmt/newsroom/featured-stories/hps/advanced-digital-twin-tech-helps
17. Red Hat (2019). “What Is Event-Driven Architecture?” Red Hat - We Make Open Source Technologies for the Enterprise, 27 Sept. 2019, https://www.redhat.com/en/topics/integration/what-is-event-driven-architecture