Waste-to-value & direct air capture · overview · Sept 2026
Design software for electrochemical reactors
Before anyone pours concrete, the reactor has already run ten thousand times.
MESSAI turns a waste stream, or a cubic metre of air, into a calibrated prediction of what comes out, what it costs, and which single measurement would shrink the uncertainty most. We sell the design and techno-economic layer. Partners build the hardware.
One engine, two front doors
Both product lines share the same core: route the input to the right physics, predict with calibrated uncertainty, and price the plant with a measurement plan attached. That shared core is the asset. The two front doors are just different inlets.
Traction you can audit
Every figure below is a static snapshot from the platform's own verification docs, dated. The DAC interval is wide on purpose: the model reports it instead of hiding it, and it names the measurement that closes it.
How money comes in
Land with a paid study, expand to the engineers who then use the tool daily, and grow to the enterprise platform once a customer has more than one site. Comparable process-simulation vendors sit at $50–150K per enterprise seat.
Why it holds
Next six months
Each milestone has a number attached so both of us know whether it happened.
What your stream could become
Pick the row that matches what you discharge or emit. Each row is a modeled system with a published baseline, and each prediction comes with an interval and an honest flag when your stream sits outside the data we have.
| Your input | System | What comes out | Notes |
|---|---|---|---|
| High-COD organic wastewater | MFC + anaerobic digestion | waterelectricitybiogas | Brewery, dairy, food processing. MFC is the most data-rich system class in the corpus. |
| High-COD, want hydrogen | MEC + anaerobic digestion | waterH₂biogas | 0.6 V applied; hydrogen instead of power. |
| Phosphorus-rich stream | MFC + struvite side-stream | waterelectricitystruvite (P) | MgCl₂ dosing; 90% P recovery in the baseline model (a 200 mg P/L centrate). |
| Ammonia-rich stream | MFC + NH₃ stripping | waterelectricityNH₃ (N) | N-fertilizer feedstock; 80% N recovery in the baseline model (a 600 mg N/L stream). |
| Heavy-metal contaminated | MFC + electrochemical polishing | waterelectricitymetals | Mining, plating, electronics effluent. Electrocoagulation removes metals into the floc sludge — removal, not recovery. |
| Any of the above, today | Conventional activated sludge | water only | The incumbent baseline every case is priced against. |
| Ambient air | Bipolar-membrane electrodialysis DAC | CO₂ streamcompressed CO₂ | Presets at 0.15 t/yr bench, 1 kt/yr pilot, 100 kt/yr commercial. Same core, different inlet. |
| Mine ventilation methane | Methanotrophic biofilter | CH₄ oxidized | Early model, not yet calibrated. Ask if relevant. |
What a feasibility study looks like
Six to eight weeks from stream data to a screening-grade answer with its uncertainty stated. The output is not a brochure; it is the number, its interval, the plant economics, and the one or two bench measurements that would move the answer most.
What we will tell you before you ask
- The DAC model has not been calibrated against a running plant. Its own anchor scorecard is 6/10 at pilot scale, and it reports specific energy 69% above the 1,000–1,500 kWh/t band published developers claim. That gap is real physics (a carbonate solvent costs two Faradays per CO₂, not one), not a bug we hide.
- Membrane life in the presets exceeds anything published. The longest reported bipolar-membrane run is about 1,100 operating hours; the presets assume far more. Membrane price and life together dominate the cost interval.
- Six of eighteen DAC cost inputs are literature-sourced; twelve are assumed. Each is labeled. A partner's quote replaces an assumption with a fact.
- Waste-to-value predictions are strongest for MFC systems, where held-out coverage was measured. Sparse archetypes carry a visible out-of-distribution flag.
Partners are the calibration. The first studies are priced accordingly.
Start with one stream
Send a flow rate and a lab sheet. Within a week you get a routed archetype, a first interval, and a scope for the study.