# About us

## We build the judgment layer for the agent era.

Who we are, the three roles we are hiring for, and how to reach us.

Canonical: https://super-intelligence.inc/about

## We build the judgment layer for the agent era.

Super Intelligence Inc. makes AI agents work the way your company decides, and makes your organization better at deciding with every outcome. We start with growth, through [MegaHit](https://megahit.ai), and expand one vertical at a time.

## Help us solve problems nobody has solved yet.

We hire people who want the hardest version of the problem. Each role below starts with the questions you would own, because that is the job.

### Designer — human–agent interfaces

**The problems you’ll own**
- How do you show a person a thousand agent actions in two minutes without hiding the one that matters?
- How do you make an AI’s decision readable at a glance: its evidence, the precedent it followed, what it rejected, how sure it is?
- What does “act, ask, never” look like, so a busy executive sets it correctly the first time and trusts it afterwards?
- When should the system interrupt a person, and when should it stay quiet?

**What you’ll do**
- Design the daily brief, the decision record, the authority controls and the token ledger, from first sketch to shipped product.
- Prototype at working fidelity and test with the operators who run agent teams.
- Own the company’s visual language, from the product to the figures in our research.

**You might be a fit if**
- You have shipped dense, data-heavy products such as developer tools, analytics or financial software.
- Information design is your strength, and you can defend a design decision with evidence.
- You would rather build the prototype yourself than describe it.

### AI Architect — the management layer

**The problems you’ll own**
- How do twelve agents share one picture of the world without drowning each other in context?
- What consistency model fits a shared state that many agents write to at once, and what happens when two of them commit plans that cannot both be true?
- How do you store a decision so it can be replayed faithfully years later, against the definitions that were believed at the time?
- How do you grant authority in tiers and keep every automated action reversible?
- How do you tie every token to an outcome, across agents we did not build?

**What you’ll do**
- Own the architecture of the management layer: shared state, context routing, the decision record, authority and rollback, cost accounting.
- Define the interfaces that let any vendor’s agent be managed, reviewed and taught.
- Build the replay and non-regression harness that every learned change must pass.
- Work with our researchers to take methods from a paper to production.

**You might be a fit if**
- You have designed and operated distributed systems in production, and can talk about how they failed.
- You have built agent systems that went past the demo: multi-agent, long-running, tool-using.
- You are strong on data modelling and consistency, and you care about observability.
- You write clearly enough that others can build on your decisions.

### Researcher — organizational intelligence

**The problems you’ll own**
- How do you estimate the value of a decision policy when outcomes arrive 60 days late and are confounded by everything else the company did?
- How do you assign credit across agents that produced a result together?
- How does a system know when not to decide, and stay calibrated as the business drifts?
- How does one agent’s lesson transfer to the others without breaking what they already do well?
- When is an expert’s minute worth more than another thousand tokens?

**What you’ll do**
- Take one of the six problems in the [Super Intelligence Lab](/research) and own it end to end: formalize it, build it, measure it on real organizational data.
- Design the measurement that could prove us wrong, then run it.
- Write up what holds, including the negative results.

**You might be a fit if**
- You have a research record in one of: causal inference and off-policy methods, multi-agent learning, continual learning, uncertainty quantification, human–AI decision making, or knowledge representation.
- You implement your own ideas.
- You prefer messy real data to a clean benchmark.

## Book a demo.

**How to apply.** Write to [k@supermem.io](mailto:k@supermem.io) with the role in the subject line and one thing you built that you are proud of.

Tell us which agents you run and one decision you wish they made better, and we will show you the Management Agent on your own case.