Eduardo Freitas / Lab

Lab notebook · Santa Catarina, Brazil

Questions about living things, worked out in software.

Science projects by Eduardo Freitas, filed by area. Each entry starts with the question it asks and what came of it. How it was built comes second.

ECOarea 1

Ecology & biodiversity

Where species are recorded, and where nobody has looked.

BioFrontier SC

Question

Where in Santa Catarina has nobody looked yet?

Biodiversity records cluster along roads, around universities and wherever somebody already went. An empty patch on the map can mean nothing lives there, or that nobody has been.

Result

An open hexbin atlas of where nobody has looked yet. It ranks biodiversity survey gaps across the state, so a student or a lab can find the places most likely to hold something unrecorded.

69,290
occurrence records
7,616
species
2,396
localities ranked
Method

Records from GBIF and iNaturalist are deduplicated across sources and binned into H3 cells. Each cell gets a frontier score from survey effort, observer count and habitat, measured as native forest cover from MapBiomas. People in the field can send observations back, and a record counts once two others independently agree on it.

Read the scores as a hypothesis. The weights come from first principles and have not been calibrated against expert judgement, so the ranking says more than any single number. Each location shows whether weak, moderate or strong evidence stands behind it.

Wantedresearchers who know SC

Ecologists, taxonomists, field biologists and labs who know Santa Catarina: help calibrate the frontier score. Say whether the weights are right, and which places to check against expert judgement.

Or add observations in the atlas. A submission counts once two others independently agree on it.

Built with Next.js · TypeScript · H3 · Leaflet · Supabase

Santa Catarina drawn as 5,852 hexagonal cells. Surveyed cells glow blue, densest along the coast; much of the interior is empty outlines.
Plate 1. Santa Catarina in 5,852 cells, from the atlas's own data. unsurveyed -> most records
AGRarea 2

Agriculture

Measuring a crop closely enough to act before it is lost.

Integrated Hydroponics Farm System

Question

Can a hydroponic farm catch a problem before it costs a harvest?

In hydroponics the nutrient solution is the soil. When its conductivity or pH drifts, the plants feel it long before anyone sees it.

Result

Continuous monitoring from the sensor on the tank to the screen in the grower's hand. EC, pH, temperature and humidity are read constantly, so a problem is caught early. Presented at Latin.science, Latinoware 2025, and published in the proceedings by the Brazilian Computer Society.

  1. 01

    Controller

    C++ · ESP32 · PlatformIO

    Reads the sensors and publishes over MQTT.

    Source
  2. 02

    Ingestion

    Python · MQTT · MongoDB

    Persists the stream of readings as it arrives.

    Source
  3. 03

    API

    NestJS · Prisma · DDD

    The back end, built on clean architecture.

    Source
  4. 04

    App

    React · React Query · Tailwind

    The grower's view of the farm.

    Source
  5. 05

    Infrastructure

    Docker · Compose · Shell

    Brings the whole system up in one step.

    Source
Fig. 2. The system as five repositories, sensor to screen.
Notices
Call for collaboratorsopen

The lab is open.

If you work on any of these questions, or want a tool built for yours, write.

Supportoptional

Free, and staying free.

These projects are free and stay free; a donation pays for hosting, data and hours.

QR code for the PayPal donation page
scan to donate