Praxia Labs

We experiment before turning it into a product.

This is what we're testing. Every experiment carries its real status: nothing in Labs is sold as a finished product.

Lab · pattern already used in prototypes

Matching and decisions

Profiles, criteria and a score that explains itself.

Engines that match profiles against criteria and return ranked recommendations, with the reason written down. We've used them in investment, real estate and marketplace prototypes.

01ProfileStructured data on each side

02CriteriaWhat matters and how much it weighs

03EngineMatches every profile against every criterion

04ScoreWith the reason for each match

05RankingOrdered, without noise

06RecommendationA person decides

We don't publish the real criteria or scores of any system.

In development

Multimodal layer

One workflow, different engines.

We're exploring a provider-independent layer that picks the right engine for the task, quality, cost and availability. It's an experiment, not a finished replacement for any tool.

01RequestWhat needs producing

02RouterPicks the engine for the task

03Text

04Image

05Voice and video

06ResultReviewed before it's used

Experiment · internal use

Video, voice and image

AI content and conversation, with a person in charge.

Voice agents: María is a demo of a Praxia agent that talks by voice or text.
Short video: an in-house generator that turns a script into short video pieces, which we use for our own communication.
Image: reading receipts, delivery notes and photos with measurements, already working inside Praxia Oficios.