Eyes do not scale
One agronomist cannot walk every edge of every plot in the same week.
AI product · Agritech
Crop intelligence the field can use.
Fitocrops reads farm data and returns a recommendation on the grower’s dashboard.
5 Modules
6 Steps
5 Partners
2025 Founded
Company active since 27 June 2025. The product is in build. This site describes that work — no invented clients, hectares, or revenue.
The gap
The data is already on the farm. The decision still is not.
Fitocrops exists for that gap. Not for another report.
Files do not talk
Images, weather, machines, and notebook lines sit in different places.
A model with no face
A notebook result never reaches the person who sprays or irrigates.
No version
If last month’s call cannot be opened again, it cannot be improved.
A confident miss
A score with no “this does not fit” line is worse than silence.
The season moves
The useful answer is for this window, not for a slide deck in the off-season.
Platform
Five modules. One product.
The model is the product. The grower is the user.
Field vision
Images and field readings go in together. The model marks what needs a look this week.
Data pipeline
Ingest, clean, keep a history. Cloud Storage on the way in. BigQuery for the record.
Models
Train and version on Vertex AI. The crop problem sets what the model learns.
Inference
An endpoint the operation can call. Cloud Run or GKE, depending on load.
Grower dashboard
A recommendation on a screen. Not a notebook.
Reading
What the operation sees
A line, a version, and a log. The numbers below are a sample slip, not a live score from a farm.
Product preview
Simulation
A simulation of the recommendation — not a live farm.
Strip marked for a walk72%
Weather join attached100%
Readings that did not fit18%
Sample only. Fitocrops does not publish a client metric on this site.
A simulation of the recommendation — not a live farm.
How a reading moves
Six steps, from the field to the text the operation opens.
-
01
Field
An image, a sensor, or a note the farm already keeps.
-
02
Data
It lands in the pipeline, versioned, so you can go back to it.
-
03
Train
The crop problem becomes a model. Vertex AI holds the version.
-
04
Infer
The operation asks. The endpoint answers in that same frame.
-
05
Recommend
A line on the dashboard. The grower does not open a notebook.
-
06
Adjust
What the field did comes back. The next model starts there.
Limits
What comes back. What stays out.
The product is a recommendation with a boundary. The boundary is part of the delivery.
On the slip
- The plot cut you sent
- What the model marked, with the version
- The window — this week, this pass, this plot
- What this reading left out
- A clear miss when the data does not fit
- The words the operation already uses
Not on the slip
- A replacement for the agronomist
- A yield promise
- A price list printed before we see the operation
- A client logo we do not have
- A hectare count invented for the homepage
- A notebook the grower has to open
Intake
What can enter the pipeline
If the farm already keeps it, the pipeline is built to take it — with a source.
Image
Phone, drone, or a camera the operation already uses.
Weather
Station or grid. Joined to the plot, not floating in a tab.
Soil
A sample or a map the farm already paid for.
Machine
Passes, rates, and the trail the equipment writes.
Plot map
The boundary the operation already draws.
Notes
The line someone wrote after walking the field.
Season
Crop, stage, and the dates that frame the reading.
Yield
What came off, when the farm already records it. Fuel for the next version, not a trophy.
Crops
The crop sets the model.
Soy, rice, corn, wheat, pasture, horticulture. These are the readings the platform is designed to hold. Not a list of clients.
Soy
Canopy, stand gaps, and the spray window joined to weather.
Rice
Water notes and canopy, in a crop Rio Grande do Sul already grows.
Corn
Stage, machine pass, and the image of the stand.
Wheat
A short season window. Disease look plus weather.
Pasture
Cover, bare soil, and the grazing note.
Horticulture
Short cycles. Small plots. Image does more of the work.
Stack
Where the models live.
Training, storage, and serving sit in one place. Vertex AI, BigQuery, Cloud Storage, Cloud Run.
Vertex AI
BigQuery
Cloud Storage
Cloud Run
GKE
IAM
- FieldThe file hits Cloud Storage with the plot id.
- DataThe pipeline writes a row in BigQuery and keeps what it dropped.
- TrainVertex AI serves the pinned version for that crop.
- InferCloud Run returns the slip, or a miss.
- RecommendThe dashboard shows the line in the operation’s words.
- AdjustThe walk note returns and waits for the next version.
Who it is for
Farms and co-ops that already collect data and still decide by eye.
Growers and co-ops
A recommendation for this season, not an 80-page report.
Operations with data
Images, weather, plots, machines. If the data exists, the pipeline can read it.
How we charge
A subscription to the platform.
Care
Farm data stays with the operation.
A recommendation is useless if the farm cannot tell who opened the file.
IAM
Who can see which farm is a setting, not a shared login.
Source kept
Every table row can point back at the file that made it.
No resale
We do not sell farm data. The subscription is for the software.
A miss is logged
When a reading does not fit, that line stays. It is how the next version gets better.
Region with the setup
The stack is Google Cloud. The region is chosen with the operation. The company is in Brazil.
Contract before files
Sensitive images can wait. The first note does not need them.
Team
Who makes Fitocrops
A small team in Santa Maria, RS. Talk to the people who build it.
Gerson Rodrigues
Managing partnerGerson Luis dos Santos Rodrigues Junior, managing partner of Fitocrops Agroscience LTDA, in Santa Maria, RS. Master’s in agronomy, and a doctoral candidate in the same field. He brings nematology, biological inputs, research and development, marketing, and the product itself: field data in, a recommendation out.
Fabio Mulinari dos Santos
PartnerFabio Mulinari dos Santos, partner of Fitocrops Agroscience LTDA. He builds with the group in Santa Maria. The field is where the data comes from and where the recommendation goes back.
Season
The window changes. The slip changes with it.
One product, four moments. The model is pinned to the moment the operation is in.
Planting
Stand, gaps, and whether the first pass should wait for the soil.
Vegetative
Canopy and the strips a person should walk this week.
Reproductive
The spray or irrigation window, joined to weather.
Harvest
What came off, written back.
The difference
Same farm. Two ways of deciding.
Without a version
- The image sits in a folder.
- The weather sits in another tab.
- The note is in a notebook.
- The call cannot be opened next month.
- A miss looks like confidence.
With Fitocrops
- The image lands with a plot id.
- Weather is joined to that plot.
- The note comes back into the same history.
- The slip names the version that spoke.
- A miss is a line, not a guess.
A day
How a reading is supposed to move.
A sample day. Not a farm we already run.
06:40 The file arrives
Images from the edge of plot 04.
07:10 The row is written
Cloud Storage keeps the file.
07:20 The version answers
Cloud Run returns the slip, or a miss.
07:30 The dashboard opens
One line. No notebook.
11:00 Someone walks
The person decides. The model does not.
18:00 The note returns
What the walk found waits for the next version.
Ground
Drawn from the south of Brazil, outward.
The company sits in Santa Maria. This is scope, not a client map.
Rio Grande do Sul
Home. Rice, soy, wheat, corn, pasture.
Santa Catarina
Shorter fields, more relief.
Paraná
Soy and corn at scale.
Cerrado
A later version, when an operation there brings the data.
We do not draw a coverage map of farms.
Road
Where the work actually is.
Open
Company active since 27 June 2025.
In build
Five modules on Google Cloud.
Not yet
A first operation, a number we can print.
Never on this site
Invented hectares, invented clients, a badge we did not receive.
Company
Brazil. Founded 2025.
Fitocrops Agroscience LTDA. CNPJ 61.479.389/0001-94. Santa Maria, Rio Grande do Sul. Small company, active.
Questions, answered short
What does Fitocrops build?
An agricultural AI platform.
Is the product live?
The company opened on 27 June 2025. We are building the product.
Does Fitocrops replace the agronomist?
No.
Do you sell farm data?
No.
Where do you run?
On Google Cloud.
Are you in a Google program?
The stack is Google Cloud. This site does not display a program badge.
Contact
Bring the crop and the data you already have.
We answer with the fit, the gap, and whether a first version makes sense. Terms come after that, not before.