Services

What we build,and how we work.

Custom platforms, applied AI, regulated systems and geospatial data, from the first specification to production and the years after it. AI does a specific job inside the product, and a person stays in charge of the result.

Web platforms, mobile apps, internal tools and multi-tenant SaaS, from the first specification to production and support.

  • Web and mobile
  • Multi-tenant SaaS
  • Internal tools
  • Long-term support

01 / How an engagement runs

Working software early.
A team that stays.

  1. 01

    Scope

    We start from how the work is done today: the spreadsheet, the paper form, the existing system. You get a written specification and a plan you can hold us to.

  2. 02

    Build

    Short cycles, with working software you can use from the first weeks.

  3. 03

    Launch

    Production deployment, data migration and training for the people who will use it.

  4. 04

    Run

    We stay on for support, monitoring and the next version. We run our own products the same way.

02 / Principles

Five rules,
in every product.

  1. 01

    The AI never writes the final record

    In every product we have built, the AI drafts, suggests or detects, and a person approves before anything is recorded.

  2. 02

    Compliance is part of the first design

    Audit trails, verified identities and official sources are designed in from the start, together with the rest of the product.

  3. 03

    Each customer's data stays separate

    Multi-tenant products keep every company, venue or office isolated, down to what the AI is allowed to retrieve.

  4. 04

    Secrets stay on the server

    API keys for AI services never reach the browser. The browser only gets short-lived, limited tokens.

  5. 05

    Built for the people who use it

    We model software on the spreadsheet, the paper form or the habit people already have, so there is little to learn.

03 / Where the AI stops

Product by product,
the same line.

In every product we have built, the AI does a specific job and a person makes the final call. This is where that line sits in each one.

AI Tree Inventory

AIDetects trees in the orthophoto and grades each one by confidence.

The engineer approves, rejects or corrects every tree before it reaches the register.

Notary Copilot

AIAnswers citizens' questions from official legal documents, and shows its sources.

The notary decides. The assistant gives orientation, never legal advice.

Jehona

AIProposes a monthly plan of posts and drafts each one in the brand's voice.

The team approves the plan and checks every post before it goes out.

Staydesk

AITurns a plain request into a proposed reservation.

Staff check the pre-filled booking and save it themselves.

Zana AI

AIDrafts lessons and videos, and answers students by voice.

The course owner reviews everything before a student sees it.

04 / Where we're comfortable

Work where the rules matter.

Compliance is part of the first design: audit trails, verified identities and official sources, designed in with the rest of the product.

  • 01Fiscal systems
  • 02Accredited inspection
  • 03Legal information
  • 04Regulated investments

05 / Technology

The stack we ship on.
Chosen for the job, not the trend.

Front end
React, TypeScript, Next.js, Vite, Tailwind CSS
Mobile
React Native with Expo, installable web apps
Back end
Node.js (Express, NestJS, Fastify), Python (FastAPI, Flask)
Data
PostgreSQL, pgvector, Prisma, SQLAlchemy, Redis, ChromaDB
AI and ML
Anthropic Claude, OpenAI (incl. real-time voice), Gemini, PyTorch, retrieval, function calling
Geospatial
GDAL, PDAL, rasterio, GeoPandas, QGIS, Leaflet
Real time
WebSockets, WebRTC, web push
Infrastructure
Docker, nginx, Cloudflare, Hetzner, Azure, Terraform, GitHub Actions

Have something to build?
Tell us about it.

Start a project