Abu Dhabi · AI-to-machine automation infrastructure
A20Core builds the infrastructure that lets AI operate real-world systems — sovereign-grade, AI-agnostic, from the edge device to the national control layer.
Agritech & Food — Our flagship vertical — irrigation, controlled-environment growing, food manufacturing.
A model can reason about a pump, a line, or a field. It cannot reach one. Everything between the reasoning and the machine — the sensing, the protocols, the permission to act — is infrastructure that mostly does not exist yet.
That layer is our whole business. We are not a model vendor and not a point-solution app: we build the ground AI stands on when it operates something real, and we build it so a nation can own it.
Flagship vertical · why food, first
of farmers across the Middle East still run their farms on paper and memory — traditional, non-software documentation.
farmers worldwide manage their land with no agricultural apps or farm-management software at all.
of harvest value is lost after harvest on farms that run without intelligent systems or software.
The primary reason is a lack of field intelligence.
We build the intelligence layer the field never had.
How we work
Most infrastructure was never designed for a machine to operate. We connect AI to the plants, lines, and controllers already running, so it can read from them and act on them safely — without a rip-and-replace.
Where the existing infrastructure cannot be made AI-ready, we build infrastructure that is — designed from the ground up to be operated by a model rather than retrofitted to tolerate one.
The system
Sense is where AI meets hardware. Platform is where AI meets operations. One Sense node governs one site; the Platform governs every node, across one sector or many.
The fleet
All sites · all sectors
The monitor-and-control layer: one operating picture across every site and sector, with AI-optimised ERP, CRM, and finance modules built in — so business logic and operational logic live in the same governed system the AI acts through.
The node
One site · one asset cluster
The edge node: an optimised MCP server — Model Context Protocol, the open standard that lets AI models call real tools and systems — sitting where AI meets machinery. It speaks the machines' protocols, translates between real hardware and the models above it, and executes automations locally, so a lost link does not mean a lost site.
Orchestration · unified ops · embedded ERP / CRM / Finance
A20Core Sense
MCP edge · site 1
A20Core Sense
MCP edge · site 2
A20Core Sense
MCP edge · site 3
Platform modules
Know your people
Employee management, visas & documents, KPIs and incentive tracking.
Know your farms
Production insight, harvest tracking, predictive planning, yield tracking.
Know your customers
Customers, distributors and partners — tracked in the same governed core.
Know your profits
P&L, cost prediction and revenue estimation, connected to real operations.
The agent team inside — each one specialised, each one governed
Live farming assistance from models tuned on agricultural knowledge.
Function-calling for IoT — runs pumps and climate under guardrails.
Vision models on the field — CLIP & Grounding DINO watching crops.
Large-data query and analysis, translated into human language.
Agents on live hardware
Through Sense, the agents don't just read about the farm — they operate and monitor the live hardware. A sensor speaks, an agent reasons, a function executes on the real valve, and the result is verified and logged. Every step below is from a real overnight scenario.
3 a.m. Soil moisture on block A-07 reads 18% — below the crop's threshold. The node reads the sensor directly and raises the signal to the agent, rather than waking a person to go look.
What the agents can call
read_soil_moisture(block)read_climate(zone)set_pump(line, state, mins)set_ventilation(zone, %)raise_alert(issue, severity) Every function is scoped, rate-limited, and logged — operators stay in command.
Illustrative · not live data
AI architecture
Their approach
"Jack of all trades — master of none."
An assistant that talks about the farm.
Our approach
A team that operates the farm.
Tayeb AI 2.0
Agronomist agent
The industry treats every AI as an LLM — all looks, no substance in the field. A chatbot cannot run an irrigation line.
Agricultural-context models — AgriCLIP (developed at MBZUAI) and Grounding DINO — reaching 99.994% plant-detection accuracy on phenotyping datasets, and on average 24% better at agricultural detection than fully fine-tuned general models.
One smart AI is not enough. Each agent is good at one job — a coordinated team of specialists, each governed, each accountable.
Monitor · Control · Analyze · Explain
Interfaces
Minimum input, maximum output: no complicated forms standing between a farmer and the job, while the heavy interpretation happens on our servers — not in the farmer's head.
For the farmer · mobile
Multi-lingual, optimised for touch, built around big buttons and plain words — usable with gloves on. Record the work and get back to it.
For the manager · desktop
The full analytical dashboard — monitoring, control, insight and planning — with the embedded agent team watching for the blind spots a busy manager can't.
Let us handle the data — you handle the harvest.
What we own
Any one of these can be bought somewhere. Holding all three — and making them answer to each other — is the part that is hard to assemble after the fact.
Purpose-built for sensing and control in real operating environments — heat, dust, distance, and bad power included.
Our own servers for MCP — the open Model Context Protocol that connects AI to real tools — let models create and run automations against real systems, not merely observe them.
The validation, safety limits, and data-residency controls that make autonomous action acceptable to a sovereign operator.
Where it runs
A vertical is a packaged application of the same Platform and Sense system — not a separate product, and not a change of company. Agritech and food security is where we are deepest and where our government-facing work concentrates today.
Flagship vertical
Agritech & food security
Roadmap · 2029–30
Water
Roadmap · 2029–30
Power
Roadmap · 2029–30
Industry & production
Who we serve
National food-security oversight: every farm in the network, one picture. POC-first adoption, with data residency inside the sovereign boundary — sovereignty and residency are non-negotiable.
Producers running many farms and assets — hydroponics to open field — who want unified monitoring, control, and cross-site AI optimisation.
Our infrastructure under a partner's own offering — an affordable farm manager from the smallest plot to the multi-hectare super farm.
The moat
The world's infrastructure wasn't built for AI. Ours is.
Sequencing over dates
2024
Farm-ops MVP live on real UAE farms — control, monitor, plan.
2025
Platform takes shape — first Sense edge nodes bridge AI and machinery.
Now · 2026
ADGM registration underway · government POCs · ~60-person build-out across AI, hardware, platform and data.
2027–28
Primary agri-operations layer for the UAE — a national farm data lake with real-time signals.
2029–30
The same infrastructure applied to water, power and industry — first GCC expansion.
2033
Open APIs — an international exchange for AgTech and infrastructure data.
Indicative roadmap — we commit to the order of the work, and stay honest about the dates.
Leadership
Chief Executive Officer
CEO of Agri Nova Investment & Al Samarat (UAE)
Leads two operating companies in the Emirates' agriculture and investment landscape — he runs the farms this platform was built on.
25 years in security systems & technology
A career spent making critical, real-world systems dependable — the discipline A20Core now applies to AI.
Active investor across agriculture & technology
Domain depth in the flagship vertical, and the network that opens its doors.
~60-person build-out underway · C-suite + four technical divisions · Abu Dhabi
Ministries, authorities, and operators: tell us what you run today and what you want it to do on its own. We will tell you honestly whether we are the right layer for it.
hello@a20core.comAbu Dhabi · United Arab Emirates