RaiOS · the Rai platform
Test thousands of crosses before you plant one.
Pilots practise thousands of flights in a simulator before they fly. Breeders can do the same. RaiOS reads each line's DNA, breeds virtual generations thousands of times, and grows the best candidates in Thai weather and soil over about 30 past seasons. You plant the few with the best odds, and you see the risk before you do.
Accounts are by invitation during the pilot. Self sign-up is coming.
01 · The knowledge layer
One knowledge layer under every answer.
Most crosses never produce a winner, genome-only tools stop at the parents, and one forecast hides the bad years. What changes that is evidence lined up across genes, places, seasons and practice. Every prediction, simulation and piece of advice below draws on it.
Updated daily · counted 31 Oct 2026. Counting began 31 Oct 2026; growth shows once there's a week of daily counts.
Public records Rai holds and lines up, including sources whose licence is still under check. They are not validated training examples and say nothing about accuracy.
First week of counting 24 Oct 2026 to 31 Oct 2026
Crops with the most new records
- Maize+69,412
- Soybean+2,510
- Pearl millet+60
- Cowpea+12
Fastest-growing layers
- Management+185 %
- Environment+88.3 %
- Phenotype+3.7 %
New public sources
- All sources+17
- Datasets15
- Papers2
New pests and photos
- Pest and disease entries+1,013
- Photos+984
Much of the store was assembled during this first week, so these are start-up numbers, not a weekly rate. Growth of the public records Rai holds and links. It never reflects who uses Rai or what they ask.
02 · Predict
What will this line be like?
Upload a VCF or PLINK file of your own line. Several AI models read the DNA and predict its traits; a final step combines their answers, because no single method is best for every trait. You also see the closest known varieties, and the settings the rice growth model needs for the line.
Runs today Your files stay private to you, and deleting an upload removes everything worked out from it.
03 · Cross, virtually
Which cross is worth making?
For each pair of parents, the simulator shuffles their chromosomes the way nature does and breeds the offspring forward to a stable line. It repeats this thousands of times, because every cross turns out a little differently. The answer isn't one guess but a frequency: how often this cross gives an outstanding, stable line.
Runs today Shown here: two blast-resistance genes brought into RD6. RD6's own DNA isn't public, so a KDML105 line stands in, as the result says.
04 · Grow, virtually
Will it thrive here, and how often?
The best lines are grown in a rice growth model, day by day, from sunlight, temperature, rain, soil water and nitrogen, at each target site over about 30 past seasons. You see how often the variety meets your must-haves, how steady it is from year to year, and the risk at flowering. Your breeders make the final choice.
05 · Advise
One field, worked out for this season.
- Every layer of the plot. Satellite greenness, the soil map, your lab test, this season's rain, the forecast and haze. See the layers →
- A farm chat. One conversation per farm, kept across seasons. It remembers what you tell it, and you can change or forget any of it.
- Fertiliser for this field. The amount that brings this soil to the crop's ideal level, plus what the crop takes away. Published tables appear only as a comparison.
- Options over many seasons. Fertiliser, water and planting-date options for one field, compared over many past seasons.
- Your lab report, read. Upload the PDF as it is; every value shows its label and unit as printed, and nothing counts until you confirm it.
06 · The crop world and your data
Open to explore. Private when it's yours.
The crop world is the public face of RaiOS: every crop drawn as an island, each opening into its knowledge graph. It is opening to everyone, and your own uploads stay yours.
Explore without an account
Crops, varieties, places and sources, readable by anyone.
Sign up yourself
A free account, a farm or a project, no invitation needed.
Upload your own data
DNA files and lab soil reports. Private by default; never added to shared data or used to train shared models unless you agree in writing.
07 · How sure
How RaiOS differs from genome-only tools.
| Question | Genome-only tools | RaiOS |
|---|---|---|
| What it answers | Which parents to cross | Which parents, where the variety grows best, and how to farm it |
| Environment | Not modelled | Thai soils, weather and seasons |
| Farm management | Not modelled | Fertiliser, water and planting-date options per field |
| Proof | Stops at a prediction | Tested first on your past trials, then on monitored plots with partners |
| Quality | Mainly yield and agronomic traits | Lab-measured grain quality and valuable compounds coming |
The credibility test comes first
Before we propose a single new cross, we run RaiOS on your past trials and ask it to predict results you already know. If it can't predict the past, it shouldn't be trusted with the future. You judge the result.
Run the test on your past trials →Your data stays yours
Your data is held in Thailand, used only for the work we agree, and stays yours. Results are published jointly, with you as a named partner.
Book a demo
Bring one question you're working on.
We'll run it with you and show you what RaiOS can and can't answer today. For agencies, we start by predicting your past trials.
- A short call with the people who build Rai
- Your crop, your province, your question
- What Rai can answer today, and where it can't yet
Or write to hello@everyrai.ai
The crop world, open to everyone.
Explore Rai's crop world without an account, sign up yourself, and bring your own data, private unless you choose to share it. Join the waitlist and we'll tell you the day it opens.