About us
Infinity Nova is the name under which we build vertical software platforms for regulated sectors and maintain our own line of R&D and advanced artificial intelligence research. Two people lead the team, and that scale is a decision: it lets us answer for every line of code and every claim we publish.
Activity
Development of software platforms for regulated sectors, R&D and advanced AI research.
Legal form
Company undergoing incorporation. Until it is registered, ownership of the assets and responsibility for the project rest with the natural persons behind it.
Base
Carlet, Valencia, Spain.
Intellectual property
Proprietary licences declared on the assets of value. The code is not published, and its authorship and integrity are evidenced by cryptographic signing of the development history.
Asset inventory
A dated internal catalogue with an eight-point profile and declared maturity per asset.
The eight pillars
The eight pillars are a deliberate transposition of the principles of the Montessori method — the person as protagonist, the guide who does not impose, the prepared environment, the self-correcting material, mixed-level groups, freedom of choice, rounded training and coexistence — to the bond between a professional and the system that assists them. The pedagogical framework is not ours and we say so; applying it to the design of software for regulated sectors is.
Whoever uses the system directs their own path: they choose where to start, what to go deeper into and in what order, according to the problem in front of them. The tool imposes no single itinerary and forces nobody through steps that add nothing to that case.
The system observes, suggests and prepares the ground; it does not replace the judgement of whoever decides, nor interrupt more than necessary. It proposes; the person disposes and signs.
An ordered, legible space adapted to the real work, where what is needed is at hand and what is incidental stays out of the way. The aim is for a professional to reach their result without asking for help or memorising the manual.
Every piece of the system carries its own verifier, so an error becomes visible the moment it is made and not three steps later. Whoever is working detects and corrects it themselves, without depending on an external review.
The newcomer and the veteran coexist in the same tool, and that is deliberate: the system can be used on the surface and also in depth. What the expert configures, the newcomer benefits from without having to understand all of it.
The sequence and the pace are set by whoever is working, not by the software. A system that forces an order alien to the real task ends up filled in carelessly, and then produces false data that looks like good data.
Our models are trained on the practice of the trade and not only on theory: what is done, in what order, under which regulation, and what counts as a good result in that sector. It is the difference between a model that recites and one that is useful.
Person and system adjust to each other: the tool learns from how it is used, and whoever uses it gains room as they get to know it. When the two disagree, the person prevails and the disagreement is recorded, because that is where you learn what needs fixing.
One precision that is not minor: none of the eight attributes intention, emotion or judgement of its own to the machine. The system proposes and the person signs, here and everywhere else in this house.
How we work
These are not the values on the poster by the front door. They are rules that have already changed specific product and research decisions, and several of them have cost us time or made us discard finished work.
In a team that punishes error, error gets hidden; and what gets hidden eventually surfaces in production, in front of the client. We record what goes wrong with the same care as what goes right. That is why we publish research findings that qualify our own hypotheses instead of filing them away.
A system that forces a team to change how it works does not get adopted: it gets worked around, filled in carelessly, and ends up producing false data. We start from the real task, not from the org chart or the feature that happened to look good.
Artificial intelligence proposes; a person with a name answers for the decision. That boundary is written into the governance of our systems, not left to the good judgement of whoever is operating them at three on a Tuesday afternoon.
An interface that demands too much memory does not fail on the quiet day: it fails during the incident, in a hurry, with someone watching. We count the steps, the context switches and what the user has to hold in their head, and we treat it as a requirement like any other.
We do not sell what is not built, we do not promise deadlines we cannot hold, and we do not take on work that would require claiming more than we can demonstrate. Every asset profile states its real maturity, including the parts that are not yet comfortable.
The Team does not compete on number of deliveries. They compete on understanding a domain to the bottom and on sustaining what was delivered years later. We work at a pace that allows reviewing, documenting and auditing our own work, because in regulated sectors that is the product.
Client data does not leave its perimeter unless it must. Models run where the data is, not the other way round.
Systems decide and act, but always under human supervision, with an emergency stop and no ability to widen their own permissions.
Every relevant decision records who requested it, why and what changed. If it cannot be audited, it is not deployed.
When something breaks, the system closes rather than letting traffic through. Availability is never bought at the cost of security.
Every asset publishes its real state: solid base, functional, or specification. We prefer a correct expectation to a brilliant demo.
Cryptography, models, automation and interconnection are shared. No vertical is rebuilt from scratch.
Every asset in the portfolio is published with its real status: documented architecture, automated tests and verifiable internal audit. Where phases remain to be completed we say so in the asset profile, because we prefer the declared maturity to be the same that any due-diligence process will find.