We are building AI models that replace the need for AI in your workflows.
Gravitas is the deterministic execution layer for organizational workflows.
CampusWay, our browser extension for university portals, is the front end that proves it. The execution layer is the business. Known requests run as reviewed workflows with zero AI calls. A reasoning model handles only what has not been mapped yet, and Nomos, our map-building model, is in training to turn those runs into new maps.
Two layers, one company.
The split is deliberate. CampusWay earns trust and builds the map at no cost to students: read-only, draft-only, every action logged. The engine underneath it, meaning the maps, the runtime that replays them and the models that keep them current, is what we will license to institutions and partners.
| Dimension | CampusWay Front end | Gravitas engine Back end |
|---|---|---|
| Role | Proof and validation | The core business |
| Built for | Students, then university staff | Institutions, enterprises and partner firms |
| Business model | Free to students; builds the map and the workflow data | Recurring licenses and white-label revenue |
The proof is specific and measured. Here is what runs today.
Where we are.
Internal benchmarks, October 2026.
- 97
- campus workflows across university sites and portals, each a reusable asset matched from thousands of phrasings.
- 57/57
- public mapped tasks completed end to end, 100%, with zero AI calls.
- 5/5
- form drafts filled and stopped before Submit. The student presses the final button.
- 0
- AI calls on a mapped run. It takes about 3 seconds and follows reviewed steps.
On real portals. CampusWay works on public university pages and inside signed-in student portals, and rolls out in invited student cohorts, each one gated on safety and reliability.
A request nobody has mapped yet goes to a reasoning model once, under a strict action policy. After review, it becomes a mapped run like the ones above. Closing that gap across every campus is what our models are for.
Three models, three costs to remove.
Every dollar this system spends on AI goes to three places: working out what a request means, solving a task nobody has mapped, and recovering when a step breaks. Each of our models takes one of those away. Argus is trained and benchmarked; Talos and Nomos are in training.
Argus
Request router
Trained and benchmarked
Removes the routing call.
Argus decides which workflow a request means, or refuses it, without calling a reasoning model. It is fine-tuned from an open encoder and distilled to an 8.4 MB table, small enough to run in the browser.
- 100%
- precision on held-out, human-written requests it routes
- 0
- wrong routes on requests it should refuse
- 8.4 MB
- the whole model
- 0.16 ms
- median time to decide
Talos
Run overseer
In training
Removes reasoning on failures.
Talos oversees every run and answers with one of a small, fixed set of commands: continue, pause, stop, retry an alternative, escalate for help, or switch a failing workflow version off everywhere. A broken step is repaired from known alternatives before anything falls back to reasoning.
Its safety veto can block a final click, such as submit, pay, send or enroll, and can never allow one.
Nomos
Map builder
In training
Removes repeat reasoning.
Nomos turns successful reasoning runs into draft workflows, proposes new phrasings and repairs, and builds a graph of each school's site. Once a map is approved, the next identical request costs $0 in model calls.
Nomos proposes; people approve. The model brings the speed; human review brings the trust institutions need.
Each model makes the engine cheaper on the next campus than on the last. That is what lets one engine travel.
The expansion path.
Six steps, one engine. Each step keeps the maps, the runtime and the models from the step before; only the front end changes. The last step is the one that changes the economics.
Students. Now
CampusWay. A free browser extension that maps the campus, one reviewed workflow at a time.
University staff.
The same map becomes an internal automation layer for the offices that run those systems.
Other institutions.
Mapped structures carry across similar campuses, many of which run the same platforms.
Companies and government agencies.
Organizations that need deterministic execution and locally hosted systems.
Financial institutions and law firms.
Accounting, compliance and other work where consistency is the product.
The partner layer.
Our models and setup services, licensed to AI implementation firms, automation firms and tech service businesses. They white-label the service on the back end and deliver it to their own clients.
Why the partner layer matters.
Distribution. Each partner brings a client base we would otherwise reach one sale at a time.
Leverage. Partners deploy the engine, so we don't staff every implementation.
Reach. Some organizations hold back from generative AI because they need repeatable results and a record of every action. Reviewed, deterministic workflows answer exactly that requirement.
The market, and what the capital builds.
Students are the wedge; the buyers of deterministic process automation are the market. Founder-funded to date, we are raising to build the back end: the engine, our models and the pilots that prove them.
- $14.6B
- Global intelligent process automation market, 2024.
- $44.7B
- Forecast for the same market in 2030.
- 19.0M
- Students enrolled in U.S. degree-granting institutions, fall 2023.
Grand View Research, Intelligent Process Automation Market Report, 2025–2030 (2024). National Center for Education Statistics, Digest of Education Statistics, Table 303.10 (2024).
Where the capital goes.
| Workstream | Proof it delivers |
|---|---|
| Compiler and runtime Nomos: turning runs into maps, plus the replay runtime | Approved runs become maps that replay with zero model calls |
| Compute and evaluation Argus training and evaluation; Talos repair | A measured cost per task, and broken steps recovered without reasoning |
| Pilot delivery Invited CampusWay student cohorts | Evidence that students come back, and how fast a campus onboards |
| Security and commercial readiness FERPA school-official agreement, HECVAT Lite, WCAG 2.1 AA self-audit with VPAT, SSO | A pilot package built for university procurement |
Targets: three campus pilots and a first paid deployment. The investor materials set out the plan in detail.
Investor access.
Materials are available to accredited investors under confidentiality: company overview, product demonstration, market model and financials. Use the form, or write to us directly at the address beneath it.
Request investor materials.
This page is for general information only and is not an offer to sell or a solicitation of an offer to buy any security. Any offering will be made only to accredited investors, as defined in Rule 501 of Regulation D under the Securities Act of 1933, through definitive offering documents. Securities have not been and may not be registered under the Securities Act. Statements about future plans reflect current intentions and are subject to change. Market figures are third-party estimates and are not Gravitas forecasts.