ThermoReef

• Inspiration:

  Coral reefs are experiencing unprecedented mass bleaching due to marine heatwaves. Marine scientists and conservation teams often lack the tools to quickly synthesize disparate satellite sea-surface temperature data with real-time field reports to act before it's too late.

• What it does:

  ThermoReef monitors ocean thermal stress anomalies using NOAA Coral Reef Watch data and analyzes them with NVIDIA Nemotron models. It detects bleaching risks early and generates structured emergency intervention protocols (such as shading schedules, biological triage, and monitoring alerts) for local conservation groups.

• How we built it:

  The system ingests satellite metrics and observational reports, routing complex text analysis, entity extraction, and recommendation generation through NVIDIA Nemotron (Nano, Super, Ultra) hosted on Nebius Token Factory. The user interface provides an interactive map and risk dashboard.

• Challenges we ran into:

  Structuring real-time satellite data alongside unstructured ecological field reports to ground the LLM's reasoning and avoid hallucinations in critical ecological advisories.

• Accomplishments that we're proud of:

  Designing an end-to-end pipeline that moves beyond passive monitoring into practical, actionable reef-saving workflows using Nebius-hosted Nemotron models.

• What we learned:

  Leveraging lightweight models for fast extraction and larger Nemotron models for multi-step ecological synthesis via Nebius Token Factory.

• What's next for ThermoReef:

  Integrating live citizen-science dive logs, expanding micro-climate hydrodynamic simulations, and piloting real-world alert dispatches with marine protection areas.

Built With

• nebius
• nvidia
• python