Lua Vision bets on uncertainty-aware AI to conquer critical sectors
2–3 minutes
An AI that prefers not to take risks
Brazilian startup Lua Vision has developed Genesys PI, an AI model designed for banks, hospitals, and law firms. Instead of providing a potentially incorrect answer, the system opts to state that it lacks sufficient information, aiming to prevent harm in financial, medical, or legal decisions.
A model built to recognize its limits
David Kang, CEO, explains that the technology was created with the principle that a model must understand its own limitations. Genesys PI assesses its internal confidence, verifies the answer, and when certainty is low, responds, “I don’t know.”
We never wanted to be the next OpenAI or Anthropic. We built this for businesses and regulated markets, with one obsession from day one: a model that understands its own limits.
Paulo Câmara, CTO, notes that development took 34 months and was inspired by neuroscience research on how the human brain handles uncertainty.
Funding and proprietary architecture
The company raised $4.2 million in equity from founders, advisors, and computing credits provided by labs. The system’s architecture was built from scratch and remains protected as intellectual property, while the underlying theory is publicly available.
Business strategy and pricing
Genesys PI can be installed on a client’s servers or in the cloud, with an annual license starting at R$1 million. There is no token-based charging; costs are tied to installed capacity and required infrastructure, which remains the contractor’s responsibility.
Eronides Junior, head of sales, emphasizes that predictable spending is crucial for companies. “We charge for the license and installed capacity. Since it runs on the client’s infrastructure, charging for data traffic doesn’t make sense.”
Implementation and compatibility
Deployment follows three stages: needs assessment, configuration, and adoption. The process typically takes at least 60 days, depending on complexity. The package includes support for local or cloud infrastructure, security recommendations, and management tools.
Lua has developed its own inference technology, enabling the model to run on devices with chips from suppliers like AMD and Apple, without relying on a single vendor.
Mission and market vision
Lua aims to bring an AI solution to every Brazilian company, combining specialized technology with a pricing model that doesn’t scale with adoption. To achieve this, costs must be predictable and aligned with the value generated.
Frequently asked questions
What does Lua Vision offer? AI models for businesses, with local or cloud deployment and management tools.
How much does it cost to hire? Implementation and licensing start at R$1 million per year; infrastructure costs are separate.
Does Lua charge for tokens? No. Pricing is based on annual license and installed capacity, with no token charges.
How long does implementation take? The minimum time is 60 days, varying by project complexity.
For whom was Genesys PI developed? For Brazilian companies in regulated sectors, focusing on specialized applications.
