Ruben Sandoval Davila does not describe himself as an AI optimist or an AI skeptic. He describes himself as the person who decides where the algorithm stops and the specialist starts. That distinction, in his telling, is the entire product.
The Architect Behind AI in Patient Care Decisions
Sandoval is the co-founder and CTO of Avena Technologies, the company behind Avena Health, one of Latin America’s largest digital nutrition and wellness platforms. He directly oversees technology strategy, AI and machine learning system design, product architecture, and a twelve-person cross-functional engineering team. His title says CTO. His function, as he explains it, is closer to something like chief boundary architect.
“Healthcare regulation caps how care can be delivered and monetized even with AI,” Sandoval said. “So the advantage shifts to specialists who maximize efficiency within those constraints and can quickly adopt new technologies as regulations evolve.”
Building Clinical Boundaries Into the Platform
That framing shapes every architectural decision the company makes. In a market where most competitors build AI features and bolt them onto existing clinical tools, Sandoval’s approach has been to embed intelligence directly into the workflow so that the practitioner never has to decide when or how to use it. The system decides. The specialist works. The boundary between the two is maintained in software, and it is Sandoval who draws it.
The distinction matters because getting it wrong is expensive in both directions. Too much automation, and the product loses the human element that keeps patients engaged. Too little, and the product becomes a glorified electronic health record, replicable by anyone with a competent engineering team. The competitive variable, Sandoval argues, is not the AI itself. It is the judgment about where to deploy it.
Why Experts Need Systems, Not More AI Tools
“Experts don’t need more tools,” he said. “They need integrated systems that turn their expertise into acquisition, conversion, and retention.”
That design philosophy has produced a platform that now serves enterprise clients whose names carry institutional weight. As listed on the company’s website, Novo Nordisk and PepsiCo are among Avena’s clients, and Bayer selected the company for its acceleration program.
Sandoval’s external-facing role extends beyond the product. He has served as the company’s technical representative to Google, where Avena was selected for the Google for Startups Accelerator for Latino Founders, and where Sandoval has spoken as a partner speaker. Apple featured Avena as its App of the Day and included it on its Hecho en Mexico curated list. Google placed Avena on its own Hecho en Mexico featured apps program. Dual recognition from both major mobile platforms is unusual for any company. For a health technology platform operating primarily in Latin America, it signals a level of product quality and technical execution that earned editorial attention independently from both ecosystems.
The engineering team Sandoval leads is twelve people, cross-functional, and organized around a set of architectural principles rather than a feature backlog. The difference is meaningful. A team organized around features builds what the product manager asks for. A team organized around architectural principles builds what the system needs to maintain its logic as it scales. Sandoval’s team operates on the second model, and the decisions about where AI is deployed and where it is not are made at his level, not delegated to individual engineers or product leads.
What Happens When AI Boundaries Are Drawn Wrong
The question that surfaces repeatedly in conversation with Sandoval is what happens to the product without him. The answer, as he describes it, is not that the company stops functioning. It is that the product drifts. Without the person making the boundary decisions, the engineering team builds features. Features accumulate. The product becomes tool-heavy and workflow-light. Adoption among specialists drops because the cognitive load increases, and the platform begins to look like every other clinical management system on the market.
“The failure mode is not that the product breaks,” he said. “The failure mode is that it becomes generic. The AI gets bolted on instead of embedded, and suddenly you’ve built something that adds friction instead of removing it. At that point, you’re competing against every EHR in the market, and you have no differentiation.”
The Architecture Behind Avena’s AI Approach
Sandoval’s background shaped this way of thinking. He built Avena’s technical infrastructure from its earliest stages, making decisions about data architecture, model integration, and workflow design that became load-bearing as the platform scaled. The Constraint-Based Nutrition Engine, which enforces clinical parameters at the data layer rather than relying on practitioner review, came out of his early work. So did the End-to-End Clinical Loop, which structures every consultation into a data pipeline feeding downstream actions without manual re-entry, and the workflow integration layer that determines at each step whether automation or human intervention is the better path forward.
These are not features documented in a product spec that any competent team could reimplement. They are compounding architectural choices, each one constraining and enabling the next, rooted in a specific point of view about what clinical software should do and what it should refuse to.
The enterprise clients lend weight to this argument, though they do not settle it entirely. The enterprise clients lend weight to this argument, though they do not settle it entirely. Novo Nordisk and PepsiCo are among the organizations that have engaged the platform, reflecting a level of institutional confidence in its architecture and execution. Whether these relationships prove that Sandoval’s boundary-drawing is uniquely valuable or simply that the platform is well-built is a question that depends on how much credit you assign to architectural decisions versus execution by the broader team.
Sandoval, characteristically, does not make the distinction. He is an engineer describing engineering decisions. The broader claim, that the person who decides where AI stops and the human starts is the most important function in a health technology company, is one he states as a technical observation. He does not elevate it into philosophy. He does not need to. The product, as it operates today, makes the argument for him.
Whether the rest of the industry arrives at the same conclusion on its own, or requires a few expensive failures first, is a question Sandoval leaves to others.
