About VEXKIO
Affective computing
that respects the subject.
VEXKIO began as two separate codebases — an R&D engine called Alma and a product layer called Veskio. In early 2026 we merged them into a single platform: one SDK, one taxonomy, one monorepo, thirty-two products.
What stayed the same is the thesis we started with: the industry defaulted to streaming video to the cloud, and that is both an engineering mistake and an ethical one. Every product we ship is an argument against that default.
What we build on
Four engineering pillars.
Every feature, every product, every customer conversation traces back to one of these four decisions. They are the substrate of the whole platform.
Edge-first by default
Inference happens on the user's device. We designed the feature vector to be small enough that what leaves is obviously non-biometric — twelve numbers per second, nothing else.
One shared taxonomy
Five classes — neutral, apertura, fricción, tensión, desconexión — run across face, voice, sign language and fusion. A single vocabulary across the whole platform.
Consent as UX, not paperwork
The SDK enforces consent prompts. No "we collect data" fine print. The subject sees, in plain language, what leaves the device and what doesn't.
Compliance is engineering
GDPR Art. 9, Colombia's Ley 1581, CCPA, HIPAA — we architected the system so compliance is a property of the data flow, not a quarterly audit scramble.
How we got here
From research bench to platform.
2025 Q1
ALMA engine starts
Core biometric pipeline — MediaPipe landmarks, voice DSP, 12-feature windowing, five-class classifier. The research bench that every later product stands on.
2025 Q3
Edge SDK ships
@vexkio/sdk hits 5.4 KB gzipped, enforces on-device inference, and lands in the first private-beta call center.
2026 Q1
Platform consolidation
Alma Engine and Veskio merge into a single VEXKIO root. Thirty-two products, one SDK, one taxonomy, one monorepo.
2026 Q2
Public launch
Marketing site, docs, dashboard, pricing — and the first paying verticals: call centers, retail, telemedicine.
How we work
Three non-negotiables.
The subject keeps the frame
We do not ship full-resolution video to cloud inference, even if the customer asks. The whole platform is built on that line.
Open about models
Taxonomy definitions, bias audits, model cards — published openly, versioned in git, referenced from the product.
Clarity over cleverness
Dashboards, API contracts, prompts. A non-technical buyer should be able to read a VEXKIO report and know exactly what it does and doesn't claim.
Team
Small team. Opinionated platform.
VEXKIO is built by a focused team of ML engineers, full-stack developers, designers and a legal advisor embedded from day one. We deliberately stayed small — thirty-two products on one SDK is only possible if the whole team shares a single mental model of the platform.
We publish model cards, open issues against our own taxonomy, and keep the product roadmap public. If you want to know what's next, the answer is in the repo.
At a glance
- Founded
- 2025
- Consolidated
- 2026 Q1
- Products
- 32
- SDK size
- 5.4 KB gzipped
- Compliance
- GDPR · CCPA · 1581
- HQ
- Distributed · LATAM / EU
Want to work with us?
Two ways in.
Start a free trial if you want to build with the SDK today. Talk to sales if you need enterprise compliance paperwork, on-prem deployment, or a white-label agreement.