Infrastructure for environmental stewardship.
Not just software for utilities. Water is the resource every other system runs on. The way we operate it has to outlast any one platform.
Built by operators, for operators.
AquaSys was started by people who had run water assets before they ever wrote software for them. That ordering matters. The product is built around the operator’s day, the regulator’s audit, and the public’s quiet expectation that the water turns on. Everything else follows from that.
Three commitments.
Operator first
The system is designed by an operator council that includes utility directors, control room leads, and dam safety engineers. The roadmap is set in the open. Any feature that adds work to the operator’s day without removing more work elsewhere does not ship.
Open hydrology data
Where customer data ownership allows, anonymised hydrology baselines are contributed back to the public scientific record. Operators benefit. Researchers benefit. The next utility in the same watershed benefits.
Climate resilience as a design constraint
Models are built to be retrained as climate baselines shift. There is no single calibration that lasts. The platform assumes the baseline will move and treats that as an engineering requirement, not an edge case.
The founding charter.
Written down in 2024 and revised once since. Two pages. It sets out what the platform will and will not do, what the company will and will not build, and how the operator council reviews both.
- 01
The operator stays in authority.
AquaSys is built as a decision-support layer, not a decision-replacement layer. The system never writes back to the control plane. Every recommendation is reviewable, reversible, and attributable. The human on the loop is the human in authority.
- 02
Every answer is cited or it is not given.
No Syfter response is returned without a Citation Path back to the source document, the PDF page, or the sensor ID. If the system cannot ground an answer, it says so. We treat hallucination as a defect, not a feature.
- 03
Customer data stays with the customer.
Sovereign deployment is the default. No customer data is required to leave the customer environment. Federated learning is opt-in, scoped, and auditable. The vector store lives inside the VPC.
- 04
The roadmap is set in the open.
An operator council of utility directors, control room leads, and dam safety engineers reviews the roadmap quarterly. Any feature that adds work to the operator’s day without removing more work elsewhere does not ship.
- 05
Climate baselines are assumed to move.
Hydrology and dam-safety models are designed for continuous recalibration. No single training run is treated as permanent. Retraining cadence, data lineage, and drift metrics are surfaced in the platform itself.
- 06
We will not build what we will not run.
Anything we ship to a customer, we run against our own reference deployments first. If we would not put it next to a regulator’s audit log, we do not put it next to yours.
Drafted 2024 · Revised 2026 · Reviewed quarterly by the operator council