Skip to main content
Suquia Bytes
Español Contact

Product studio · Córdoba, Argentina

We build software, hardware and games end to end

Three engineers with one shared obsession: the curiosity to work at the edge. We design, build and ship. We started with our own.

Let's talk See what we built

What we have already built

  • Orisen

    AI-powered smart cane for blind people

  • Wyrdvow

    Narrative mobile RPG on a deterministic engine

  • This background

    Hand-written WebGL shader, no libraries

What we do

Three verticals. Each one backed by something we have already built and that you can inspect.

  1. Software

    Product from zero to production: mobile, backend, infrastructure and applied AI.

    We take on the whole product, not a slice of the ticket. Technical design, development, deployment and whatever comes after. We integrate AI with engineering judgement: swappable adapters, versioned prompts, fallbacks and a clear line between what the model generates and what the system decides.

    • React Native · Expo
    • NestJS · Prisma
    • PostgreSQL
    • Cloud Run · R2
    • Nx

    The proof

    Orisen

    Expo app, NestJS API on Cloud Run, Python bridge and an in-house training pipeline.

  2. Hardware

    Firmware, BLE and WiFi connectivity, and the app that drives it.

    This is where there is least competition and where we already have something working. When the JavaScript runtime is not enough we write the native module: that happened with the cane network transport, and we solved it by opening the socket from the operating system.

    • PlatformIO
    • Raspberry Pi
    • BLE · WiFi
    • Native Expo modules
    • Binary protocol

    The proof

    Orisen

    Cane firmware, our own binary protocol and BLE/WiFi session orchestration.

  3. Games

    Game development with the same discipline we apply to software.

    A game with AI inside is only testable if the rules are reproducible. That is why we separate narrative generation from the engine: the AI writes, the deterministic engine decides. The same run with the same seed always produces the same result.

    • Deterministic engine
    • AI adapters
    • WebGL · GLSL
    • Expo · React Native
    • Zod

    The proof

    Wyrdvow

    And the background you are looking at: a 4 KB hand-written WebGL shader, with no library at all.

Work

We have no wall of client logos. We have our own products: we designed them, we built them, we maintain them. That is considerably harder to fake.

Software Hardware In active development

Orisen

AI-powered assistive technology for blind people

The problem

Mobility aids detect obstacles, but they do not describe the environment. Orisen adds AI-assisted perception on a connected cane and communicates it through audio.

What we built

  • Mobile app in Expo / React Native
  • NestJS + Prisma backend on Cloud Run
  • Python bridge running on a Raspberry Pi Zero 2W
  • In-house model training pipeline
  • Cane firmware in PlatformIO
  • Nx monorepo with independent deploys per application

One technical decision

The phone connects to the cane over WiFi, but Android decides on its own which interface each socket goes out through: the channel kept dropping even when preflight came back green. We wrote a native Expo module so the operating system opens the socket, and kept our binary protocol in TypeScript. Native transport, typed protocol, one API for both iOS and Android.

Three-dimensional render of the Orisen smart cane: a white cane with blue light guides along its body and a glowing spherical sensor at the tip.
Product render of the Orisen cane.
Games In development · playable end-to-end MVP

Wyrdvow

Narrative mobile RPG with AI

The problem

An RPG where the character's oath alters rolls, progression and ending. The narrative is generated with AI, but the rules cannot be arbitrary: they have to be reproducible.

What we built

  • Expo 55 app with React Native 0.83 and React 19
  • NestJS 11 API with Prisma 7 and PostgreSQL
  • Deterministic, pure, tested rules engine
  • Runtime contracts shared between app and API
  • Swappable AI adapters and Cloudflare R2 storage
  • Admin dashboard in Vite / React

One technical decision

We separated narrative generation from game rules: the AI writes, the deterministic engine decides. The same run with the same seed always produces the same result, and that is precisely what makes a game with AI inside it testable.

Screenshot of the Wyrdvow app: a narrative scene screen in dark theme, with health and stress bars, the scene text and two decision options flagged as risky.
A Wyrdvow narrative scene in the mobile app.

What money cannot buy

We build for people who cannot see the screen. This site too.

Orisen exists so a blind person can move around with more information. It would be incoherent if the page that tells that story could not be navigated with a screen reader.

  • Fully keyboard navigable, with our own visible focus ring
  • Contrast ratios computed as numbers, not eyeballed
  • Heading hierarchy without skips and landmarks with accessible names
  • All motion switches off under prefers-reduced-motion
  • Works without JavaScript, contact form included

Audit us. Run this page through axe, through Lighthouse or through your screen reader. If you find something broken, write to us and we will fix it.

How we work

There are three of us. You talk directly to whoever writes the code, with no layers in between. It also means rigour is not optional: we have no spare people to fix what went wrong.

  1. Contracts before code

    Runtime schemas are shared between app and API. When a contract changes it breaks at compile time, not in production.

  2. Pure domain logic

    Rules live isolated from infrastructure and are tested without a database or network. The Wyrdvow engine is deterministic by design.

  3. Explicit monorepo boundaries

    Every library declares what it may depend on. Architecture is verified in CI, not in code review.

  4. Decisions in writing

    Decision records, hardening plans and documented scope before touching the code. We write it for ourselves and it ends up serving the client.

  5. We go down to the metal when needed

    If the runtime is not enough, we write the native module. If the library weighs 250 KB to draw a background, we write the shader.

  6. No scope surprises

    Scope agreed in writing before we start. If something changes we discuss and quote it; it does not show up at the end on the invoice.

Who we are

Three engineers, no layers in between

No sales team and no management. The person who answers your email is the one writing the code. What connects us is the curiosity to work at the edge.

  • 01

    Ignacio Castro

    Engineering · Co-founder

    Bio coming soon.

  • 02

    Lucas Ribotta

    Engineering · Co-founder

    Bio coming soon.

  • 03

    Joel Logarzo

    Engineering · Co-founder

    Bio coming soon.

Let's talk

Tell us what you want to build. You get an answer from one of the three people who would work on it.

It helps us tell you quickly whether we can do it. It is not a fixed quote.

What you want to build, by when, and what exists today.