By industry
AI, IoT and software for agriculture and controlled-environment growing
Farms and growing operations already generate more data than they use: sensor readings, input purchases, labour, yields and sales, each in its own place. We build the layer that puts them in one record, flags what is drifting while it can still be corrected, and keeps the agronomist in the decision rather than replacing them with a confident model.
Two weeks to a written build plan · no obligation to build with us
- Commerce we built for a growing brand
- HydroponicsCommerce we built for a growing brand
- Events survive poor connectivity
- Field-gradeEvents survive poor connectivity
- Growing decisions stay with a person
- Agronomist signsGrowing decisions stay with a person
- To a written build plan, via the audit
- 2 weeksTo a written build plan, via the audit
Where the time goes
If two of these are true, there is something here
Written the way the people doing the work describe it, not the way a process document does.
Three spreadsheets and a WhatsApp group
Yield in one workbook, inputs in another, sales in a third, and the operational detail that explains all of them in a chat thread that scrolls away.
Sensors were installed and then ignored
Moisture, EC, pH, temperature — all being recorded, none reaching the person who would act on them, and no alert when a reading goes out of range overnight.
Cost per unit is a guess
Inputs, labour and energy are not attributed to a block, a cycle or a crop, so the profitable lines and the loss-making ones look the same in the accounts.
Traceability is reconstructed after the fact
A buyer or an auditor asks which inputs went onto which block in which week, and the answer is assembled from memory and receipts.
What we connect to
The systems you are already running
Agriculture spans very low-tech and very high-tech in the same business. The integration job is usually to make a sensor network, a phone in a worker’s hand and an accounting package produce one record.
- Sensing and control gear
- Soil and EC / pH probes
- Climate sensors
- Irrigation controllers (read-only)
- LoRaWAN
- Weather station and forecast APIs
- Operations
- Field data apps (Flutter)
- Offline capture
- QR / barcode lot tracking
- Labour logs
- Commercial
- QuickBooks
- Xero
- Tally
- Storefront / marketplace channels
- Buyer reporting formats
What we build here
Five things worth building for agriculture & agritech
Each one names the systems it touches, because an integration that is not named is an integration nobody has thought about yet.
- 01
One record per block or cycle
Sensor readings, applications, labour hours, harvest weights and sales joined to the block and the cycle they belong to, so a question about a crop has one place to be answered from.
- Touches
- Sensors · field app · accounting
- What changes
- Yield, cost and quality can finally be compared across blocks and seasons.
- 02
Alerts while there is still time to act
Out-of-range readings and trends compared against each block’s own baseline and its plan, escalated by how fast the situation is moving, to the person on shift rather than to a dashboard nobody is watching.
- Touches
- Sensor feed · alert routing · field app
- What changes
- Problems are caught overnight instead of found in the morning.
- 03
Field capture that works without signal
Applications, observations, labour and harvest recorded on a phone at the point of work, offline, syncing when the device gets back in range — with photographs attached where a note is not enough.
- Touches
- Flutter app · offline store · central database
- What changes
- Records are made where the work happens, not reconstructed at a desk.
- 04
Cost and margin by crop
Inputs, labour and energy attributed to blocks and cycles and set against realised sales, so cost per kilo is a number rather than an instinct.
- Touches
- Purchase records · labour logs · sales ledger
- What changes
- Crop mix decisions are made on measured margin.
- 05
Traceability and buyer reporting
Lot-level history from input to harvest to despatch, with the evidence attached, exported in the formats buyers and certification schemes ask for.
- Touches
- Lot records · application logs · export templates
- What changes
- An audit request is a report rather than a week of reconstruction.
How it is put together
How a reading becomes a decision an agronomist owns
Agriculture is the sector where an overconfident automated decision is most expensive, because the feedback loop is a season long. So the system watches and escalates; the agronomist decides.
- 01
Capture
Sensors, field app entries and purchase records, all timestamped and tied to a block. Field entries are offline-first, because connectivity in a growing area is not a reasonable assumption.
- 02
Normalise
Units, calibration offsets and sensor drift handled explicitly. An uncalibrated probe reading treated as truth is how a system loses an agronomist’s trust in the first week.
- 03
Compare to plan
Each block has a plan — target ranges, expected stage, planned applications. Deviation from that plan is what gets attention, rather than raw values against a generic threshold.
- 04
Escalate
Routed by severity and rate of change to whoever is on shift, with the readings and recent history attached. Slow drift gets a daily note; a failing pump gets a call.
- 05
Record the decision
What was seen, what was decided, who decided it and what happened afterwards — which is what turns a season of readings into something the next season can learn from.
Your data
What we will and will not do with it
Stated plainly, and without claiming a certification we do not hold. If a requirement here is harder than what is written, say so on the call and we will tell you whether it is buildable.
We do not drive irrigation, dosing or climate control
We read controllers and sensors and we recommend. Writing setpoints to a dosing or irrigation system can destroy a crop in hours, and that authority stays with your control system and your people.
No agronomic advice from a model
Nutrient, spray and treatment decisions are made by qualified people. The system surfaces deviation and history; it does not tell you what to apply. Generic model advice in this domain is confidently wrong in exactly the places that matter.
Yield prediction is offered honestly or not at all
Useful prediction needs several seasons of your own data. In year one you get measurement, comparison and alerting — which is worth having — and prediction becomes a conversation once the history exists.
Proof
We have built for a growing business already
Gardyn is a US hydroponics brand. We built the storefront, catalogue and fulfilment flows behind it, sized for a seasonal demand curve rather than an average week. Add the ReCyrcle side of our work — IoT sync to field hardware with events that survive bad connectivity — and you have both halves of what a growing operation needs.
- Live storefront
- US east coastLive storefront
- Built to hold the curve
- Seasonal peaksBuilt to hold the curve
Technology
What this gets built with
We pick the boring option unless there is a reason not to — the framework is the part most likely to be abandoned before your system is.
- Sensing
- LoRaWAN
- MQTT
- Modbus
- Weather APIs
- Calibration handling
- Field software
- Flutter
- Offline-first sync
- QR lot tracking
- Photo capture
- Data
- TimescaleDB
- PostgreSQL
- Block / cycle model
- Grafana
- Commercial
- Accounting integrations
- Cost attribution
- Traceability exports
- Next.js dashboards
How the engagement runs
Five phases, each with something you can hold
Every phase ends in a named deliverable. You can stop after any of them and keep what has been built.
- 01Week 1–2
Sit with the work
We watch the process as it is actually done, not as the process document says. The output is a map of every step, which system it touches and how long it takes.
DeliverableProcess map and a shortlist of what is worth automating - 02Week 2–3
Design the access layer
Which systems get read, which get written to, what scope each credential carries, and which decisions stop for a person. Agreed in writing before anything is built.
DeliverableArchitecture and permission model - 03Week 3–8
Build against your real data
Built in your sandbox against real records, not a demo dataset. The write path goes through a review queue from the first day, so nothing lands in a live system unseen.
DeliverableRunning system in staging, with the review queue - 04Week 8–10
Run it beside the humans
It runs in parallel with the people doing the job. We compare the two outputs, fix what differs, and only then start letting the routine path through unattended.
DeliverableAgreement rate report and a go-live threshold - 05Ongoing
Hand over and watch
Source code in your repository, deployment in your accounts, cost and failure dashboards you can read without us. Support if you want it, not because you are locked in.
DeliverableRepository, runbook and monitoring
Why Devs Core
Three reasons that are checkable
We run two of our own products
Cadenza and Budge are ours. What an agent costs per run, where it fails and which decisions need a person came from operating them, not from a vendor deck.
The write path is reviewed by default
Anything that changes a record in a system of record goes through a queue a person can see. You choose what graduates to unattended, and when.
You own the code
Source in your repository, infrastructure in your accounts, model keys in your name. There is no runtime of ours you have to keep paying for.
Questions
Agriculture & AgriTech — the questions we get asked
Can AI tell us when to irrigate or feed?
It can tell you that a block has drifted from its plan, how fast, and what happened the last time it did. What to apply stays with your agronomist. The difference matters: the first is measurement, the second is advice we are not qualified to give and would not want automated.
Our sensors are from three different vendors.
That is the normal case and it is most of the integration work — units, sampling rates and calibration conventions all differ. Normalising them into one model is exactly the layer worth building, and it is where the value shows up first.
We have no connectivity in the growing area.
Then everything is offline-first: the field app records locally and syncs when it gets back in range, and sensors buffer at the gateway. This is the same design we needed for collection hardware on the ReCyrcle work.
Can you predict our yield?
Not credibly in the first season. Prediction needs several cycles of your own data including the bad ones. Anyone promising accurate yield prediction from a fresh install is selling you a model fitted to somebody else’s farm.
We also sell direct to consumers. Can that be part of the same system?
Yes, and it should be — that is where cost per unit meets realised price. We have built the commerce side for a hydroponics brand already, and joining it to production records is what makes crop mix a measurable decision.
Is this affordable for a mid-sized operation?
The measurement and alerting layer is, and it is usually where the payback is. Vision-based grading and prediction are larger projects that should wait until the basic record is trustworthy — we would rather sequence it that way than sell the expensive part first.
Related
What agriculture & agritech buys most often
- Read more
Custom Software Development
The operational system underneath — the one the packaged software did not cover. React and Django, built to be handed over.
From $14,000 - Read more
AI Integration Services
The access layer between the software you already run and any AI model — so an agent can reach your ERP, CRM or helpdesk without being handed the keys.
From $8,000 - Read more
Mobile App Development
iOS and Android from one Flutter codebase, or native where the hardware demands it.
From $9,000 - Read more
Computer Vision Development
Detection, inspection and counting from camera feeds, running on the floor rather than in a research notebook.
From $9,000
- By industry
IoT & Connected Devices
The devices work. Getting what they know into the software does not.
- By industry
Sustainability & Recycling
Collection data sits in hardware nobody can get a report out of.
- By industry
E-commerce & Retail
Two people do nothing but re-key orders between the store and the ERP.
Tell us which process is costing you.
Thirty minutes, no preparation, no deck. You describe what keeps going wrong and we tell you whether AI is the answer — including when it is not.
Prefer email? contact@devs-core.com