Multiversal VenturesOPEN SOURCE · APACHE-2.0
OpenReading
The unified interface for every reading. One schema across fifteen-plus OCR and document backends. Every provider fails somewhere; route around it.
AGENT WORKFORCE
Copy and paste your advantage.
An agent workforce for regulated industries.
We build working agents that do a defined job, in your tools, in your words, under your rules. They shadow before they act, and they say what they will not do.
See the workforce in action
From tracing an unpaid claim to learning your team's workflow. Watch how an agent works in your tools, with your rules and your review.
We build and operate full-stack AI inside regulated industries.
Healthcare, insurance, lending and tax: the places where the data is sensitive, the rules are heavy, and trust is the product. We are not a model vendor or a demo shop. We build the agents that do the work, prove they are safe, and keep them compliant as the rules change. Then we run them. Multiversal is a permanent owner-operator, revenue-funded, built to still be here in ten years.
01 · STUDIO
The studio funds the work
We take a workflow that used to need people and own it end to end, from the research to the running system. You keep the IP.
Studio engagements →02 · PRODUCTS
The products compound it
Problems we were hired to solve more than once became products: mortgage intelligence, bank-data recovery, reimbursement, the payer rulebook, the data estate.
Solutions →03 · LAB
The lab keeps both ahead
Hammer Labs builds world models of regulated systems and publishes the research. It ships into production, and it gets published.
World models →The shelf
Everything on offer, in one table. No price list: terms are set per engagement, in a working session.
| # | Offering | What it is | Who it is for |
|---|---|---|---|
| 01 | Studio engagementsSTUDIO | Full-stack AI built and run inside a regulated workflow, on a scoped statement of work. | COO, CTO, chief AI officer in healthcare, insurance, lending, tax |
| 02 | AI Growth EngineSTUDIO | AI engineers we hire, train, equity-align and manage inside your team. | VP Engineering, CTO building an AI team |
| 03 | Agent workforceSTUDIO | A working agent that performs a defined job role, in the tools where the work already happens. | Owners and operators with a job nobody has hours for |
| 04 | Hammer LabsLAB | World models of regulated systems, so software is tested against the rules. | Banks, insurers, payers, health systems, lenders |
| 05 | FeneroPRODUCT | The intelligence layer for mortgage lending: income, credit, DTI, compliance. | Loan officers, brokers, credit unions, lenders |
| 06 | BrookPRODUCT | The exception handler for Plaid. Failed bank connections recovered, same schema. | Fintech lenders gated on a bank connection |
| 07 | Coverage IntelligencePRODUCT | The living map of where a healthcare product gets paid, kept current. | Digital health, medtech and diagnostics leaders; investors |
| 08 | GAVELPRODUCT | The payer rulebook as software: coverage, criteria, payment, appeals. | Provider groups, MSOs, RCM platforms, device companies |
| 09 | CartographerPRODUCT | The enterprise data estate mapped in an afternoon, metadata only. | Data and AI leaders at large enterprises |
| 10 | GeneseeLAB | The benchmark for healthcare payment-integrity agents. Every leaked dollar labeled in an answer key. | Teams building revenue-cycle and payment-integrity agents; buyers evaluating them |
Proof, in production
Terabytes
Of documents read every month
Complete control over the most cost-optimal way to process each page: every page routed to the cheapest reader that gets it right. Built and hardened over terabytes of production documents a month.
OpenReading →1,000s
Of agents running in production
Under a harness, with rigor: tested against world models of the rules they operate inside before and after every release.
World models, from Hammer Labs →40–50%
Reduction in AI costs
Unit economics by design. Inference on small language models, running on CPUs and lower-tier GPUs rather than frontier APIs.
How the studio builds →$1B+
Companies we work inside
Our work runs in production at several billion-dollar companies in healthcare, insurance and lending. They are not named here: their method is their advantage. References on request.
Ask for a reference →Who this is for, and who it is not
We would rather walk from a bad-fit engagement than win it. We excel at zero-to-one, bias toward speed, and de-risk early. We lose in meeting-heavy cultures.
Lean in
A regulated market: healthcare, insurance, lending, tax, finance.
The workflow runs on documents and rules, and meaning lives in text.
Skilled people spend their days reading, checking and re-keying.
Errors carry real cost: rework, audit findings, denials, fines.
An owner with budget who can decide in weeks, not quarters.
Walk away
A committee wants a vendor to attend standing meetings.
Approval chains with no single owner and no date.
“Build us a chatbot” with no workflow and no number attached.
A demo circus rather than a working session.
AI as theater: a press release in search of a project.
What we publish in the open
OPEN SOURCE · APACHE-2.0 / MIT
Cuttlefish
Native tooling for AI agents. A coding agent hands off a job and Cuttlefish runs it locally, sandboxed, against a local model. The files stay on the machine.