Coagen
Autonomous coagulation, flocculation and chemical dosing that hits target water quality at minimum chemical cost.
Six products, one plant-edge runtime, one audit backbone, one data flywheel. Turbix is the perception every control product acts on, Coagen, Aerix and Membraq are the control pillars for dosing, aeration energy and membranes, Simflo rehearses every plan before write-back, and Permia gates all of it to stay permit-safe.
Illustrative scenario: one municipal activated-sludge wastewater plant with a membrane train, planning the next 24 hours after a rain forecast. Not a measured run.
These are the stages of one perceive, dose, aerate, filter, disinfect and comply loop plus the twin that plans it, not separate bets. They share the plant-edge runtime, the sensors and cameras, the compliance governance, the control connectors and one compounding flywheel.
Autonomous coagulation, flocculation and chemical dosing that hits target water quality at minimum chemical cost.
Autonomous aeration and biological-process control that cuts the plant's biggest energy bill while holding effluent spec.
Membrane and RO fouling prognostics with autonomous flux, backwash and cleaning control for desalination and water reuse.
Plant-edge water-quality perception: vision plus online-sensor fusion that sees turbidity, clarity, foam and anomalies in real time.
The as-run plant digital twin: simulate dosing, membranes and aeration to hit water quality, compliance and energy before the shift.
Autonomous compliance and permit assurance: predict excursions, prove every action and keep the plant permit-safe.
Autonomous coagulation, flocculation and chemical dosing that hits target water quality at minimum chemical cost.
Coagulant, polymer and chemical dosing is the most craft-bound and cost-sensitive step in a treatment plant. Operators set doses from jar tests and experience, then chase a drifting, weather-driven influent with fixed setpoints and periodic lab samples. Overdosing wastes expensive chemicals and drives sludge and residual-metal problems; underdosing breaks turbidity, TOC and filter-loading limits and risks disinfection-byproduct and permit excursions.
Coagen closes the loop from influent perception to dosing-pump action. It forecasts influent characteristics, predicts the coagulation, flocculation and settled and filtered water response, and continuously computes the coagulant, polymer and pH-adjust dose under turbidity, TOC, alkalinity, sludge, byproduct and cost constraints. It runs in shadow, assist or bounded closed-loop mode, learning from every operator correction and lab confirmation.
| Capability | Status |
|---|---|
| Influent and dose-response forecasting | Real, fine-tunable per plant |
| Multi-objective constrained dose optimization | Real, via cuOpt |
| Physics-informed priors: mass balance, alkalinity, residence time | Planned |
| Dose-change justification and operator Q&A over SOPs and permits | Real, via the model router |
| Learning from the closed-loop correction corpus | Planned |
Jetson Orin at the plant edge runs TensorRT dose-response and influent models for on-prem control when cloud links degrade. DGX or HGX trains per-cohort dose-response transformers. TensorRT and Triton serve inference with per-tenant routing and audit ids; NeMo fine-tunes the justification models; cuOpt solves the dose schedule; RAPIDS builds features from analyzer, lab and SCADA tags. NVIDIA AI Enterprise and NIM for supported self-hosted serving. Items beyond Jetson, TensorRT, Triton and cuOpt are planned.
Autonomous aeration and biological-process control that cuts the plant's biggest energy bill while holding effluent spec.
Aeration is the single largest energy consumer at most wastewater plants, often the majority of plant electricity, because blowers are run conservatively high to guarantee nutrient removal and avoid permit excursions. Dissolved-oxygen, activated-sludge and nutrient-removal dynamics are nonlinear, load- and temperature-variable and craft-controlled; fixed DO setpoints waste energy in low-load periods and still risk nitrification upsets, bulking and nutrient breaches during surges.
Aerix models DO demand and the activated-sludge and nutrient-removal state, forecasts influent load, and continuously computes the minimum-energy blower and DO setpoint schedule that still meets nitrogen, phosphorus and effluent targets. It coordinates aeration with dosing for external carbon and chemical phosphorus removal, predicts biological upsets before they breach permit, and runs shadow, assist and bounded closed-loop against blower and DO controls.
| Capability | Status |
|---|---|
| Load and DO-demand time-series forecasting | Real, fine-tunable |
| Biological-state estimation and upset prediction | Real; physics-informed priors planned |
| Constrained energy minimization of blower schedules | Real, via cuOpt |
| Twin-in-the-loop aeration simulation for pre-shift planning | Planned, via Simflo |
| Control justification over SOPs and permits | Real |
Jetson Orin runs TensorRT DO and biological models per basin for low-latency, resilient on-prem control. DGX or HGX trains the biological and aeration transformers and runs the oxygen-transfer simulation. cuOpt solves blower and DO scheduling under nutrient, permit and tariff constraints; TensorRT and Triton serve forecasting and state estimation; Omniverse and Modulus carry the aeration twin; NeMo and RAPIDS handle process-reasoning tuning and telemetry ETL. NVIDIA AI Enterprise and NIM for on-prem serving. Twin, NeMo, RAPIDS and NIM are planned.
Membrane and RO fouling prognostics with autonomous flux, backwash and cleaning control for desalination and water reuse.
Membranes and reverse osmosis are the backbone of desalination and potable and non-potable reuse, but they are dominated by fouling and scaling: transmembrane pressure and flux drift with feed water, temperature and biofouling, so operators run conservative flux, over-clean and waste chemicals, water and membrane life, or under-clean and risk irreversible fouling and integrity breaches. Cleaning timing, backwash frequency and flux setpoints are craft decisions made with lagging indicators.
Membraq fuses feed-water quality, transmembrane pressure, flux, differential pressure and conductivity to predict fouling and scaling trajectories, forecasts the flux and recovery that maximize membrane life, and schedules backwash and clean-in-place just in time to minimize energy, chemical and water use, while guarding integrity and pathogen log-removal for reuse. It runs shadow, assist and bounded closed-loop against membrane skid controls and learns from every cleaning outcome.
| Capability | Status |
|---|---|
| TMP, flux and fouling time-series prognostics | Real, fine-tunable per train |
| Physics-informed membrane models: TMP and flux, mass transfer, osmotic pressure | Planned |
| Constrained optimization of flux, recovery and cleaning schedules | Real, via cuOpt |
| Anomaly and integrity-breach detection from sensor fusion | Real |
| Justification over membrane SOPs, warranties and permits | Real |
Jetson Orin runs TensorRT fouling and integrity models per skid for resilient on-prem control. DGX or HGX trains the fouling-prognostics transformers across the fleet and runs membrane simulation. TensorRT and Triton serve prognostics and anomaly models with audit routing; cuOpt schedules backwash, cleaning and flux under energy, recovery and integrity constraints; Omniverse and Modulus simulate fouling and flux for the twin; RAPIDS and NeMo handle telemetry ETL and reasoning. NVIDIA AI Enterprise and NIM for self-hosted serving. Twin, RAPIDS, NeMo and NIM are planned.
Plant-edge water-quality perception: vision plus online-sensor fusion that sees turbidity, clarity, foam and anomalies in real time.
Treatment plants run largely blind between lab samples. Turbidity, clarity, floc and settling behaviour, foam and bulking, filter breakthrough, membrane leaks, chlorine residual, pH, nutrients and TOC are judged by periodic grab samples, delayed lab results and operator rounds. Online analyzers exist but are siloed, noisy and drift-prone, and they are not fused with the visual cues experienced operators read by eye.
Turbix fuses plant-edge computer vision on clarifier, basin and filter cameras with online analyzers into a continuous, calibrated, drift-corrected water-quality state, predicting excursions and detecting anomalies before lab confirmation. It provides the real-time perception and soft-sensor signals that Coagen, Aerix, Membraq and the compliance agent act on, and flags integrity and safety anomalies plant-wide.
| Capability | Status |
|---|---|
| Clarity, foam, floc and anomaly vision at the plant edge | Real, fine-tunable |
| Multimodal sensor and image fusion into calibrated soft sensors | Real |
| Time-series anomaly detection and excursion forecasting | Real |
| Synthetic augmentation of rare visual and sensor events | Planned, via Replicator and Cosmos |
| Analyzer drift and fault estimation | Real; physics-informed priors planned |
Jetson Orin runs TensorRT vision and fusion models at each camera and analyzer cluster, with design targets under 100 ms per camera and 15 to 30 frames per second. DGX or HGX trains the vision and fusion models on plant imagery, corrections and lab results. Metropolis and DeepStream carry the multi-camera pipelines; Holoscan fuses analyzer streams at low latency; Cosmos and Omniverse Replicator synthesize rare foam, fouling, breakthrough and surge scenarios; Triton, NeMo and RAPIDS serve, reason and build features. NVIDIA AI Enterprise and NIM for self-hosted edge deployment. Synthetic data, NeMo, RAPIDS and NIM are planned.
The as-run plant digital twin: simulate dosing, membranes and aeration to hit water quality, compliance and energy before the shift.
Operators make high-stakes control decisions, dose ramps, blower setpoints, cleaning windows and wet-weather response, with no way to test them before committing to the live plant. Design models are static as-designed simulations that diverge from the real, aged, fouled, drifting plant. Rare, dangerous scenarios are exactly where operators have the least data and the most risk, and they cannot be rehearsed on a live public-health asset.
Simflo builds a live, calibrated GPU simulation of each plant, hydraulics, coagulation and clarification, filtration, membrane fouling, aeration and oxygen transfer, and biological and nutrient response, that tracks the real plant as run rather than as designed. Before each shift it simulates the next 8 to 24 hours and evaluates 50 to 500 candidate control plans, returning the plan that hits target water quality, compliance and energy, and generates synthetic rare-event scenarios to stress-test autonomy and train the other agents.
| Capability | Status |
|---|---|
| GPU physics simulation: hydraulics, oxygen transfer, membrane flux, biological kinetics | Planned fidelity per process |
| Data assimilation and calibration of the twin to live telemetry | Real |
| Surrogate and neural-operator models for fast twin evaluation | Real; Modulus planned |
| Synthetic scenario generation | Planned, via Cosmos and Replicator |
| Control-plan search over simulated outcomes | Real, via cuOpt |
DGX or HGX and OVX run the process, hydraulic and biological simulation and many parallel plan evaluations; RTX workstations carry twin authoring and visualization for engineering firms; Jetson Orin runs distilled surrogate twins at the plant for on-shift what-ifs. Omniverse is the twin environment, Modulus the physics surrogates, Cosmos and Replicator the synthetic scenarios, cuOpt, CUDA and RAPIDS the plan search, solvers and calibration ETL. NVIDIA AI Enterprise and Omniverse Cloud for hosting. Full fidelity, edge surrogates and scenario coverage are planned.
Autonomous compliance and permit assurance: predict excursions, prove every action and keep the plant permit-safe.
Water and wastewater plants live or die by permits: drinking-water standards, discharge permits, nutrient limits, PFAS and disinfection-byproduct rules and reporting obligations across many jurisdictions. Compliance is assembled by hand from SCADA logs, lab results and operator notes, and a single excursion can breach a permit, trigger fines and endanger public health. Operators react after lab confirmation; there is no continuous, predictive, auditable compliance layer.
Permia continuously evaluates plant state against every applicable permit and regulation, forecasts excursion risk hours ahead, and gates autonomy so that control actions from Coagen, Aerix and Membraq stay permit-safe. It uses permission-aware retrieval over permits, SOPs and water regulations with enforced citations, maintains an immutable, assurance-grade audit trail of every agent and operator action, and drafts regulatory reports with full traceability.
| Capability | Status |
|---|---|
| Excursion-risk forecasting and anomaly detection | Real |
| Permission-aware retrieval with citation grounding over permits and regulations | Real |
| Conformance, non-conformance and report drafting | Real, via the router; NeMo-tuned planned |
| Guardrails, schema validation and compliance gating of control actions | Real |
| Golden-set and model-as-judge evaluation gating model and prompt changes | Real |
DGX or HGX trains the excursion-forecasting models and fine-tunes the compliance-reasoning models; Jetson Orin runs excursion-risk inference and action gating on-prem when cloud links degrade. NeMo with Guardrails governs the compliance models with citation enforcement; TensorRT and Triton serve forecasting and retrieval with per-tenant routing and audit ids; RAPIDS builds compliance features from SCADA, lab and permit data. NVIDIA AI Enterprise and NIM for on-prem retrieval and LLM serving. NeMo, RAPIDS and NIM are planned.
Every Hydrira run is an ordered, inspectable sequence. This is run_scn_0418, an illustrative scenario: one municipal activated-sludge wastewater plant with a membrane train, planning the next 24 hours after a rain forecast. Ten steps, nine automatic, one human approval on the blower setpoint move, every action written to the shift record.
Intake-and-Screen · Flow, turbidity, ammonia and alkalinity at the headworks, and a 24-hour rain forecast that raises the expected influent by mid-afternoon.
Plant orchestrator · Simflo evaluates candidate dose, DO and flux plans over the 24-hour horizon and keeps the plans that hold permit margin through the surge.
Quality-and-Anomaly · Clarifier blanket, foam and colour from the basin cameras fused with turbidity, pH, nutrient and residual analyzers into one calibrated plant state.
Dose-and-Coagulate · Coagulant and polymer dose ramp for the surge window under turbidity, phosphorus, sludge and chemical-cost constraints, cited to the plant's dosing SOP.
Dose-and-Coagulate · A bounded setpoint move inside the declared rate and magnitude limits for the dosing pump, with the rollback value recorded first.
Aerate-and-Bio · Blower and DO setpoint schedule across the basins that meets the nitrogen and phosphorus targets at the lowest kilowatt-hours the biological state allows.
Plant orchestrator · The DO setpoint change is above the plant's threshold for this loop. The process engineer on shift reviews the predicted nutrient margin and approves.
Membrane-and-Disinfect · Flux held through the surge and the next backwash moved earlier, inside the integrity envelope; the disinfection residual minimum is untouched.
Compliance-and-Energy · Every planned setpoint checked against the discharge permit, the nutrient limits and the residual minimum, with excursion risk forecast for the horizon.
Quality-and-Conformance · Inputs, model versions, predictions, limits, the approval and every write appended to the immutable shift record the plant owns.
Products are licensed per process on the Process tier and bundled on Plant and Enterprise.
| Capability | Process | Plant | Enterprise |
|---|---|---|---|
| Processes covered | 1 | All on site | All in fleet |
| Control agents | One of Coagen, Aerix, Membraq | All | All, with custom models |
| Turbix perception | As the one process, or attached | Included | Included |
| Simflo twin | Not included | Included | Included, per-archetype templates |
| Permia compliance gating | Included for the process | Plant-wide, with reporting | Multi-jurisdiction |
| Robot-and-Inspect | Not included | Where policy allows | Where policy allows |
| Autonomy stages | Shadow, assist, bounded | Shadow, assist, bounded | Shadow, assist, bounded, fleet policy |
| Deployment | Plant edge + cloud | Plant edge + cloud or on-prem | Edge fleet, on-prem or private cloud |
| Support | Design-partner support | Design-partner support | SLAs |
| Outcome pricing | Not included | Optional | Available |
Every tool invocation, argument and result is written to an immutable, human-readable log, and every reasoning step is expandable. Nothing about a run is hidden from the plant. The log below is the scenario run, not a measured one.
Goal, plan, thought, action, observation, exactly as the orchestrator would record it.
Hold effluent turbidity, nutrients and disinfection residual inside the discharge permit for the next 24 hours while taking aeration energy and coagulant use to the minimum the permit margin allows.
Read influent and the forecast; rehearse the horizon on Simflo; fuse the plant state; compute dose and DO schedules; write the dose inside bounds; route the DO move for approval; hold flux and move the backwash; check the permit; commit the record.
Rain arrives mid-afternoon and the influent will rise by about a third. Dosing has to ramp before the surge, not after, and the DO schedule has to keep nutrient margin through it.
cuopt.solve_dose over a 12-hour window under turbidity, phosphorus, sludge and cost constraints; plc.write_setpoint on the coagulant pump inside the declared rate and magnitude limits.
The dose ramp holds predicted effluent turbidity under 2 NTU through the surge. The DO schedule saves energy but is above the threshold set for the aeration loop.
Energy never outranks permit margin. The DO change goes to the engineer with the predicted nutrient margin attached, and nothing is written until it is approved.
approve.request to the process engineer on shift; membrane.schedule_backwash to hold flux and move the next backwash ahead of the surge; permit.evaluate across the horizon.
Ten steps, one human approval, all parameters inside limit on the forecast, and the shift record written. Illustrative scenario; nothing here is a measured plant result.
A plant does not go from manual to bounded control in one step. Hydrira makes the stage explicit per loop, auditable and reversible at any time, and some decisions never leave the person.
| Stage | What the agent does | What the person does | When |
|---|---|---|---|
| Shadow | Predicts water quality, fouling, excursion risk and energy; writes nothing | Runs the plant as today; compares predictions to lab, analyzer and SCADA outcomes | First phase with every design partner |
| Assist | Recommends dose, DO, flux and cleaning moves with predicted quality, energy and permit impact | Approves, corrects or rejects each move; corrections train the models | After accuracy is proven against the plant's own baseline |
| Bounded | Writes to an allow-list of low-risk loops inside declared rate and magnitude limits | Sets the limits, watches the record, pulls the loop back at any time | Per loop, after the assist-mode gate and compliance sign-off |
| Exceptions | Escalates anything above the threshold, every permit-adjacent move and every release | Decides; the operator of record is never removed from compliance decisions | Always |
A treatment plant runs around the clock on variable influent, and many utilities keep plant data on site, so perception and control inference run at the plant edge. Hydrira keeps training, simulation and optimization in the cloud, in a private cloud or on-prem. Items marked planned are on the roadmap, not running.
TensorRT-optimized time-series and vision models at the plant edge; DeepStream pipelines for clarifier, basin and filter cameras; Holoscan fusing camera streams with turbidity, pH, chlorine and nutrient analyzers at 1 to 10 Hz. Jetson Thor is planned.
Water-quality forecasting, anomaly detection, control recommendation and compliance retrieval served with per-tenant routing, model versioning and an audit id on every call. Self-hosted for utility on-prem and edge deployment. Planned.
Time-series transformers, vision models and NeMo-tuned process and compliance reasoning trained on plant telemetry, operator corrections, images, alarms, lab results, permits and SOPs. Design target: weekly fine-tuning per design-partner cohort, then monthly fleet refreshes. Planned.
Simflo runs plant hydraulics, clarifier and filtration behaviour, membrane fouling, oxygen transfer and biological response as a GPU simulation calibrated to the live plant, with Modulus surrogates for fast evaluation. Planned.
Wet-weather surges, sensor drift, foaming, fouling, filtration breakthrough and biological upsets generated as scenario variants to stress-test autonomy and train the perception and control agents. Planned.
Blower and pump scheduling, chemical dosing windows, membrane cleaning, backwash timing and fleet maintenance routes solved under energy tariff, labour, asset and permit constraints.
ETL over high-volume SCADA tags, lab results, membrane data, alarms and image metadata to build features and golden evaluation sets. Planned.
Supervised robotic sampling routes, asset-inspection rovers and materials handling across intake, treatment and disinfection, where plant policy allows robots. Planned.
The numbers Hydrira is engineered against, from the GPU utilization plan.
Design targets from the GPU utilization plan, all marked aspirational in the architecture. Nothing on this page is a measured plant result.
The three people in a treatment plant whose week changes when the process runs itself, and what each of them needs from Hydrira.
Owns compliance, reliability, energy, chemicals and cost. Wants autonomous, compliant, lower-energy, lower-chemical treatment, and a number that justifies the spend.
Owns dosing, aeration, membrane and biological strategy, the process windows and the performance. Needs recommendations they can inspect, correct and trust before any loop is handed over.
Owns permit compliance, water quality, energy and environmental reporting and the licence to operate. Needs every action cited, logged and defensible in front of a regulator.
IT/OT and critical-infrastructure cybersecurity leads approve the deployment. There are no customer quotes on this site yet; design partners will be the first voices.
Hydrira reads and writes through the systems already in the plant. No parallel historian, no new HMI, no rip-and-replace. Connectors are built against each design partner's stack first; long-tail integrations are out of scope for the first release.
Plant SCADA, historians, PLC and DCS interfaces
Tag reads at 1 to 10 Hz, guarded setpoint writes, interlocks untouched
Coagulant, polymer, pH-adjust and disinfectant dosing controllers
Dose reads, bounded setpoint writes inside declared limits
Blower controls, DO and ammonia-based aeration control
DO and blower reads, bounded setpoint writes after approval
Membrane and RO skid PLCs, TMP, flux, conductivity, integrity monitors
Reads; flux and backwash writes inside the integrity envelope
Turbidity, pH, chlorine residual, nutrient, TOC and flow analyzers
Reads; drift and fault flags
Lab information systems, jar-test and grab-sample records
Reads for labels and calibration
Fixed cameras on clarifiers, basins, filters and membrane rooms; inspection robots
Streams fused at the edge; supervised routes
Utility MES, maintenance systems, permit and reporting systems, weather
Reads; shift record and report drafts written back
Hydrira writes to control equipment in a plant that keeps a community's water safe. Every capability is scoped, every write is policy-checked, every output is cited, and every action is written to an append-only audit log the plant owns.
| Item | Scope | Status |
|---|---|---|
| SOC 2 Type I + critical-infrastructure cyber review | Cloud control plane, edge | QUEUED Roadmap, first 6 months |
| SOC 2 Type II | Cloud control plane | QUEUED Roadmap, 6 to 12 months |
| SDWA · discharge permits · EU WFD · nutrient and PFAS rules | Grounding and record-keeping | RUNNING Design basis for retrieval and the audit log |
| Per-tenant isolation | Data, memory, models | RUNNING Design rule |
| On-prem or edge-only deployment | Utilities, critical infrastructure | RUNNING Design option |
Hydrira supports your compliance records. It is not a certification, it does not replace your permit and it never removes the operator of record from a compliance decision.
One policy model, one audit trail and one set of energy, chemical, compliance and water-quality benchmarks across every plant in the fleet, with each site's permits, SOPs and process models kept private to that site.
The questions operations managers, process engineers and compliance leads ask in the first meeting.
Only at the autonomy stage your plant has set, and only to an explicit allow-list of low-risk loops inside documented limits. Every design partner starts in shadow mode, where Hydrira predicts and writes nothing. Assist mode follows, where a process engineer approves each move. Bounded write-back comes last, per loop, after the accuracy gate and compliance sign-off, and can be pulled back at any time.
No. Permit-limit setpoints, disinfection residual minimums, integrity limits and discharge release decisions are read and checked, never written. Permia evaluates every planned move against the permit set and blocks anything that would thin a margin. The operator of record stays the operator of record.
The plant keeps running on its existing SCADA, PLC and DCS control and interlocks. Hydrira is supervisory, not primary. Loss of Hydrira is loss of advice, and the gap is recorded in the shift record.
Shadow mode starts with the SCADA tags, analyzer streams and lab records you already have. Accuracy is measured against your own lab and analyzer baseline, and the first-release plan converts to assist mode only when the number is proven. The design targets assume telemetry you already collect, not new instrumentation.
Telemetry, permits, SOPs and models are tenant-scoped, and an on-prem or edge-only deployment is available for utilities. Cross-plant learning is federated and IP-isolated, planned and opt-in: patterns for similar basins, membranes and influent regimes move, raw data does not.
Not yet, and the site does not claim it. SOC 2 Type I with a critical-infrastructure cyber review is on the roadmap for the first six months, Type II for the six to twelve month window. Safe Drinking Water Act, discharge permit, EU Water Framework and nutrient and PFAS rules are what the retrieval grounds on and what the audit log is aligned to; that is a design basis, not a certification.
The one with the clearest number. Aeration energy on an activated-sludge plant, coagulant cost on a dosing train or membrane run length on an RO or MBR train are the usual wedges, because kilowatt-hours, chemical invoices and cleaning counts are already measured. Turbix is the sensing wedge where visibility is the first problem.
Two to four utilities or industrial sites instrument one wedge process in shadow mode and baseline energy, chemicals and compliance from their own records. Assist mode proves accuracy and ROI with an engineer approving each move. Graduated autonomy follows on low-risk loops. Expansion adds adjacent processes, then fleet sites. Gates, not dates.
Most plants start with Aerix on an activated-sludge basin group or Coagen on a dosing train. Tell us which number you need to move.
↑↓ navigate↵ openesc close