About MESSAI

We are building a research platform for microbial electrochemistry.

A small team building research infrastructure for a field that has spent two decades waiting for its knowledge to become computable — structured, queryable, and predictive, from literature to laboratory to scale.

Loading 3D montage…

Microbes · materials · reactors
drag to orbit

Our vision

A foundation the field can build on.

The vision

To make microbial electrochemistry as designable as any mature engineering discipline — by unifying and standardizing the world’s MES knowledge into one platform for research, development, education, and commercialization.

The mission

To give researchers, engineers, and innovators the tools to design, simulate, and predict electrochemical systems grounded in 19,000+ published papers — not a handful of remembered ones — so good ideas reach the world faster.

The field never had a science problem. It had an information infrastructure problem — and that is a problem we know how to solve.

Provenance first

Every value links to its source paper. Units and measurement conditions are attached where the paper reports them, and about half of values carry a source snippet. Where a field is missing we show it as missing rather than inferring it.

Honest uncertainty

Served predictions carry an interval where one is calibrated, and say so when it is not.

Open by default

The core MESS packages — parameters, materials, datasets — are open source and community-usable.

Built between disciplines

MESSAI lives in the gap between biology, electrochemistry, and software — by design.

From research to innovation

How a discovery becomes an industry.

The distance between a brilliant result at the bench and a system running in the world has always been enormous — not because the science is weak, but because the knowledge never became something you could compute against. We see that distance as a pipeline, and we built MESSAI to make each stage flow into the next.

H₂ Planned: validated results feed back. 01 LITERATURE 02 STRUCTURE 03 PREDICT 04 DESIGN 05 DEPLOY
01

Literature

Two decades, tens of thousands of papers

The evidence already exists — scattered across hundreds of journals in incompatible formats, un-queryable and impossible to compute against.

02

Structure

Papers become data

AI reads the papers and extracts each measurement with its source, units, and conditions where the paper reports them. Harmonized to SI and a typed ontology of 835 parameters. About half the corpus is mined so far.

03

Predict

Data becomes foresight

Per-class models turn the corpus into performance predictions. Each links to its source, with a calibrated interval where one is calibrated. Where none is, we say so.

04

Design

Foresight becomes a system

Researchers design, simulate, and sweep reactors in an interactive 3D lab — testing a hundred configurations in silico before touching a bench.

05

Deploy

A system becomes impact

Validated designs reach the world — cleaner water, green hydrogen, carbon turned to chemicals — and each result is planned to flow back to sharpen the data that produced it.

The gap

The bottleneck was never the biology.

For twenty years the science has worked. Microbes have been turning wastewater into electricity, organic waste into hydrogen, and carbon dioxide into chemicals in labs around the world. What the field lacked was not another breakthrough at the bench — it lacked the infrastructure to make those breakthroughs cumulative.

Until recently, few tools could reliably read the messy, heterogeneous text of scientific papers and turn it into structured, comparable data — not at the accuracy or cost a real platform demands. Modern AI now makes that practical. It reads the published papers, extracts each measurement with its provenance, and makes the results computable.

AI collapses the decade-long lag between a result in a lab notebook and a design an engineer can build — and that is what turns a promising science into a deployable industry.

That is how we believe the field gets the cumulative knowledge it has lacked: not by one lab out-discovering the rest, but by giving the field a shared, evidence-based foundation to compound on.

19 K+
Papers in the corpus
214 K+
Measurements extracted
835
Parameters in ontology
2,819
Causal couplings

Platform figures as of October 2026, from 19,000+ published papers.

The team

Founder, scientist, operator.

The combination this category requires: platform engineering, peer-reviewed domain authority, and disciplined research operations.

Sam Frons

Sam Frons

Founder & CEO

Information-infrastructure builder across health, agriculture, and climate. Spent six years mapping the MES field before building the platform.

Zakiya Sharpe

Zakiya Sharpe

Head of Research Operations

Runs the extraction QC and validation workflows that keep the corpus trustworthy — every measurement traced from paper to queryable record.

Dr. Lane Gilchrist

Dr. Lane Gilchrist

Chief Scientific Officer

Bioprocess engineer and faculty scientist — the peer-reviewed, bench-side authority anchoring the parameter ontology and scientific direction.

Questions

Frequently asked.

What is MESSAI?

MESSAI is the intelligence layer for microbial electrochemical systems (MES). We turn two decades of published research into structured, queryable, predictive infrastructure — spanning literature search, a typed parameter ontology, an interactive 3D design lab, and performance predictions with honest uncertainty.

What are microbial electrochemical systems?

They are reactors where microbes growing on conductive electrodes drive electrochemical reactions — generating electricity from wastewater, producing hydrogen from organic waste, or synthesizing chemicals from CO₂. Microbial fuel cells (MFC), electrolysis cells (MEC), and desalination cells (MDC) are the best-known members of a much larger family.

Where does your data come from — and can I trust it?

Every value links to the paper it was extracted from. Units and measurement conditions are attached where the paper reports them, and about half of values carry a source snippet. Nothing enters the ontology without its source, and incompatible units are harmonized to SI so studies can be compared like-for-like. About half the corpus is mined so far. MES data has enormous natural spread, so we treat scientific honesty as a rule, not a footnote.

How accurate are the predictions?

Served predictions carry a value, a unit, a source and, where one is calibrated, an interval. Where none is calibrated, we say so. Corpus-wide variance in these systems runs to four figures, so a single headline number would mislead. Method and coverage are on /proof and /methods.

Does MESSAI build reactors or hardware?

No. We are the intelligence and software layer that maps, models, and predicts these systems. We don’t build or sell the hardware — we make it far cheaper and faster to design the hardware that works.

Is any of it open source?

Yes. The core MESS packages — parameters, materials, datasets, and more — are open source and community-usable. The platform is built on the belief that a field advances fastest on shared, open foundations.

How do I get started?

Explore the research intelligence surface, try the 3D lab, or browse the parameter catalog — all public. To talk about a partnership or a pilot, email founders@messai.io.