of recorded global extinctions involved invasive alien species, alone or with other drivers
Environmental edge AI
Local intelligence for faster freshwater decisions
FLUVIUS Brain is a modular system designed to help existing field equipment screen aquatic risk close to the source, preserve an auditable record, and guide expert attention. We are implementing FLUVIUS Brain in Europe, Brazil, and Canada.
Designed for screening and decision support. Ecological and regulatory judgement stays with qualified teams.
- Stage
- Early MVP
- Evidence
- Initial lake test
- First use case
- Zebra mussel risk
Enclosure concept
The actionability gap
A signal helps only when teams know where to look next
Many aquatic monitoring workflows can indicate that a threat may be present. Field teams still need a targetable location, a time‑stamped record, and enough context to decide what to confirm.
FLUVIUS Brain focuses on that gap between a signal and a field response. The aim is to place a verifiable intelligence layer near the observation, then move only useful evidence into the team's existing workflow.
estimated annual global cost of biological invasions in 2019
Detection without operational “where” information can slow confirmation and response.
Context: Convention on Biological Diversity ↗The system
One modular layer, four clear functions
The architecture separates sensing, local inference, evidence logging, and user‑facing review so each part can be tested and improved independently.
01 / 04
Capture observations close to the water
A camera and available environmental context create the field input. The module is designed to fit boats, fixed stations, and partner equipment.
Current prototype
A documented starting point with performance validation ahead
The current workbench combines a Raspberry Pi 5 host with a physical BrainChip Akida accelerator. Model development takes place off‑device. The team has completed an early hardware integration test in lake conditions.
- Host
- Raspberry Pi 5
- Accelerator
- Physical BrainChip Akida
- Model development
- Off‑device
- Planned output
- Risk signal + evidence log
Evidence boundary. Latency, accuracy, energy use, and accelerator execution remain subject to versioned, repeatable validation. The site does not claim those results yet.
Evidence before scale
The next milestone is a repeatable proof package
The prototype has crossed the first integration milestone. The work now is to freeze the configuration, measure it transparently, and validate a narrow use case with a field partner.
Evidence in hand
- Early hardware prototype assembled
- Dataset and reproducible training pipeline assembled
- Physical Akida accelerator connected to the host
- Initial lake test and field‑development archive
Next evidence gate
- Freeze hardware, software, and model versions
- Capture accelerator mapping and execution logs
- Benchmark quality, latency, and energy against a baseline
- Run a structured pilot with acceptance criteria
AquaHacking Prairies 2025
First place and CA$20,000 in seed funding
The original Drift‑Eye Swarm concept established the team's direction: bring monitoring closer to the current and help communities act earlier. FLUVIUS Brain is the modular intelligence layer emerging from that work.
Read the AquaAction announcement ↗AquaEntrepreneur 2026
Selected for Cohort 5
The commercialization program is helping the team turn a promising prototype into a defined pilot offer, clearer validation metrics, and a practical path to customer adoption.
View the AquaAction portfolio ↗Pilot fit
For teams responsible for freshwater assets and field evidence
The strongest first pilot has one clear monitoring question, known field conditions, an expert confirmation method, and a partner who can judge whether the output improves a real decision.
Environmental authorities
Defined screening workflows for protected areas, reservoirs, and managed waterways.
Water asset operators
Marinas, utilities, and aquaculture sites with a practical inspection or maintenance pathway.
Research and monitoring teams
Field programs that need traceable observations and a configurable edge workflow.
Equipment partners
Sensor, drone, and robotics makers exploring embedded environmental intelligence.
Working commercial model
Start with proof. Grow through repeatable deployments.
The current model combines paid validation and integration work with future device revenue, recurring software or model services, and partner licensing. Pricing and willingness to pay remain hypotheses to test with customers.
- 01Defined pilotOne site, protocol, and decision
- 02Repeatable deploymentDocumented hardware and workflow
- 03Recurring serviceEvidence review, updates, and analytics
- 04Partner integrationLicensing or co‑development where useful
A useful pilot starts with one decision
Bring us a freshwater monitoring question worth testing
We are looking for partners who can provide a defined site, expert ground truth, and an honest standard for whether local AI improves the field workflow.