Arvantti-Fractal® is a registered trademark of Arvantti Labs, the intelligence solutions brand of Sequoia Space. Today I am going to show you what Arvantti-Fractal® is, how it makes decisions and the three ways in which it can operate in your company: connected to a digital twin, as intelligence applied to physical operations or as a risk calculation system to decide.

This is the route. First the challenge: why deciding is difficult when there are many variables, restrictions and risk. Then what is Arvantti-Fractal® and how to decide. Then, the most important thing: the three ways it can operate. We close with how we compare scenarios, how you stay in control, and how we get started. Each index point has a link to its section.

Important decisions in an operation almost never depend on a single variable. There are many variables at time: demand, capacity, inventories, shifts, energy. There are restrictions: equipment limits, regulations, delivery windows or budget. The objectives are in tension: lowering costs can affect service or security. And there is uncertainty: failures and changes that no one planned. When everything comes together, the decision often remains in spreadsheets and the experience of a few people, and then it is difficult to explain why one option was chosen and not another. That is the problem that Arvantti-Fractal® addresses.

Arvantti-Fractal®, registered trademark, is the decision-making engine of Arvantti Labs. It is hybrid because it combines three things. Artificial intelligence helps analyze information and explore alternatives. Deterministic flows apply rules, constraints, and validations that can be reviewed line by line. And the decision models leave the variables, objectives and assumptions written. The result is not a black box: each recommendation comes with its explanation, the assumptions it used and the risk it assumes. The specific combination of these pieces is defined in each implementation.

That's how Arvantti-Fractal® decides, no matter which of its three forms it operates in. First define the variables: what can be moved and what is measured. Then the restrictions: what can never be violated, whether physical, security, regulatory or business limits. Then the objectives and their priority. With that, he creates scenarios: the current plan versus alternatives. The risk is calculated for each scenario: how likely it is to breach a restriction and what impact it would have. The result is an explained recommendation. And the last step is always human: a responsible person approves, adjusts or rejects.

This is the central message. Arvantti-Fractal® is not just a complement to Arvantti-DigitalTwin®. It can operate in three ways. One: connected to Arvantti-DigitalTwin®, where it co-simulates the plant, analyzes predictive events, compares alternatives and recommends; the operator decides. Two: like Physical AI, that is, intelligence within the equipment itself: a chip with Arvantti-Fractal® reads the sensors and calculates on site the best value of a parameter, within the limits defined by your equipment. Three: as a risk calculation system to decide, based on your business data. In the next three slides we look at each one.

First way: connected to Arvantti-DigitalTwin®, registered trademark. Arvantti-Fractal® reads PLC signals in read-only mode and never writes to the PLC or the plant. It runs its own calculation, independent of the model that represents the operation. It projects 24-hour scenarios, the current plan versus alternatives, and the recommendation reaches the operator, who accepts or rejects it. If your plant is still a project, Arvantti-Emulator®, a registered trademark, emulates it with a virtual PLC. Example: in the demonstration scale aqueduct, base and alternative 24-hour studies are generated that the operator reviews before deciding.

Second way: Physical AI. Here Arvantti-Fractal® is not loose on a server: it lives inside the hardware, on a chip mounted on the equipment itself, on site. Let's think about a device that controls a parameter: it can be off, operating within a range or on. With the usual, deterministic algorithms, you can tell it to turn on, turn off or set a value. What you can't ask him to do is decide, based on other parameters he's reading, what the best value is within that range. This already requires a probabilistic model with deterministic functions within the same chip, which is what Arvantti-Fractal® provides. In the diagram: the equipment's sensors read other parameters, the chip with Arvantti-Fractal® resolves the uncertainty and calculates the optimal value, and the actuator applies it, always within the allowed range defined by your equipment. Thus the system optimizes and makes the best decision on the site. Each implementation is defined by the customer's equipment, control system, and security standards.

Third way: a risk calculation system for making decisions. We start from the data you already have, such as ERP, spreadsheets or event logs, and we model variables, constraints and objectives. On the left, the flow canvas: when faced with an urgent order there are two paths. Alternative A puts an extra shift on line 1 and delivers on day 3; Alternative B moves orders between lines 1 and 2 and delivers on day 2. On the right, the optimization function: for each alternative the engine calculates cost plus risk, and the best one is the one lowest on the curve; in this example, B. Below, the recommendation explained, with its assumption, and your team approves, adjusts or rejects. The values ​​are illustrative. The same applies to logistics, process prioritization or selection of initiatives and budget; the result always depends on the data, assumptions and validation of your team.

In any of the three modes, the basis of the analysis is to compare before deciding. We start from the base scenario, which is the current plan. We compare it with alternatives, which you can define or that the engine explores. We do sensitivity analysis: what happens if an assumption such as demand, a failure or a price changes. And we calculate the risk of each scenario: how likely it is to violate a constraint and with what impact. Additionally, Arvantti-Fractal® connects with the AI ​​of your choice, via CLI or MCP, to explore alternatives and explain results. The criteria and their weights are defined with your team; We do not present baseline results as guaranteed, because any results depend on the data, assumptions, and validated model.

A recommendation is only useful if you trust it. That is why Arvantti-Fractal® works with explicit assumptions: variables, restrictions and objectives are written and your team can review and change them. It is designed to leave traceability: what data, what version of the model and what decision was behind each recommendation. The human is always in the loop: the analysis informs and a responsible person decides. And you can operate on Arvantti Harness®, trademark, our intelligence governance foundation, with permissions, validation and evidence. Note: Arvantti-Fractal® is an independent line and does not require Arvantti Harness® to work.

We start in four steps. One, diagnosis: we choose a specific decision, we review the data you have and we define in which of the three modes it operates. Two, decision model: we write with your team the variables, restrictions, objectives and assumptions. Three, validation: we compare the model with historical cases or with the digital twin, if there is one, and we adjust it. Four, start-up: the engine delivers recommendations in a limited scope, with human approval and success criteria agreed upon from the beginning. With that evidence we decide together whether to escalate.

We close with an open question: what decision do you want to make with more clarity. From your answer we see which of the three modes Arvantti-Fractal® fits into: connected to Arvantti-DigitalTwin®, as Physical AI or as a risk calculation system. My contact information is on the screen: Andrés Quiceno, CTO of Engineering, [email protected], and the sites arvantti.tech, arvantti.online and sequoia-space.com.