Ars ex Machina

Technology that works in the real world

An engineering and technology company: robots, software, analytics and automation for businesses that need to automate complex processes in the real world. We design, build, deploy and test it all ourselves.

We build robots that bring in the harvest, farms that run themselves, and software that takes the manual grind out of the business.

What sets us apart is our own test site. A berry farm outside Moscow, where everything we build is tested against reality: weather, mud and a full season.

We are not consultants with a slide deck. We are the people who have held both the soldering iron and the shovel.

Values

Six rules we use to decide what to build and what not to promise.

01

Reality comes before theory

A technology is worth something only when it works outside the slide deck.

02

Understand first

We do not automate chaos. First we understand the system, its data and cause and effect. Then we build the solution.

03

Engineering honesty

We do not promise what we have not yet proven. We keep «done», «doing» and «planned» apart.

04

Complex things should become simple

Complexity stays inside the technology. To the user the system must be clear and controllable.

05

A system beats a standalone product

The robot gets stronger through data. Data through analytics. Analytics through real action.

06

AXM connects technologies into one system

Physical world, data, decision, action: one chain. Not a standalone robot, not standalone software, not standalone analytics.

AXM showroom
AXM platform — the engineering inside

Agro: robot, field, dashboard

Not one technology but a chain: the robot picks, the field reports on itself, the dashboard brings it together in one window. Each direction reinforces the rest.

Where we are now

The project is in development and not yet in full operation. Below: what is done, what we are working on right now and what comes next. The «done» column holds only what we can actually show you.

Done

Perception: the machine sees the berry

  • Perception MVP on NVIDIA Jetson Orin: the rover finds a berry, judges how ripe it is, works out where to cut and brings an arm to it — the whole loop runs on board.
  • ~38 frames per second, 24 of 24 tests passing, our own labelled datasets.
Done

Structure and sensing

  • Parametric model of Bee: layout, rocker-bogie suspension, eight arms, geometry checks — the model compares itself against the concept shown on this site, so the numbers cannot quietly drift apart.
  • Firmware for the cell and the gateway builds: adaptive polling interval, encrypted LoRa link, setup from a phone over BLE. The purchase specification is ready.
  • Telemetry receiver is live on the server, and the end-to-end test «sensor → gateway → server → dashboard» has passed.
Done

The Apiary dashboard — running on a real field

  • Land boundaries pulled from the state cadastre by parcel number; field mapping with crop blocks, tracks and infrastructure.
  • The irrigation design parsed from the document: 31 zones, 29 shifts, blocks split by the number of drip lines, fertigation recipe calculated.
  • A variety reference, thirty years of local climate, and NDVI from the Sentinel-2 satellite with a season-long history.
  • All of it is open right now — the demo on our own field.
In progress

From model to metal

  • Buying the sensor hardware, bench-testing the cell and the gateway, then testing them in the field.
  • Rover mechanics: hull shape, suspension travel, masses and centre of gravity.
  • Specifications for the Queen and the Hive station.
  • Irrigation in hardware: a valve per zone, flow and pressure sensors at the pump house, an injector with EC and pH control.
Next

The first machine that drives

  • A running Bee prototype on our own field: picking in real beds, and field statistics for the picking cycle.
  • Commands from the dashboard to the machine: tasks, sorting criteria and field mapping by RTK.
The goal

The swarm

  • Several Bees in the field, a Queen with a refrigerated body, and a Hive station swapping batteries.
  • Schedules and sorting run from the dashboard, all season long.

We have left the dates out on purpose. Until the machine is out in the field, any date would be a promise rather than a plan. Once the prototype has been through its first season, the numbers will appear here.

Automation for any business

Everything that automates our own farm works outside the field too: processes, data, programs.

24h

A fresh case: one day to a working robot

A manufacturing company, custom metalworking: 4,261 CRM deals and ~3,800 emails machine-read; the robot costs a part from its drawing to within ±6%. The breakdown lives on the Analytic.

160M

The losses it found

160 M ₽ a year lost to controllable causes — price and quoting speed. The company's data knew it all along; nobody had asked.

01

The dashboard and the program

The AXM dashboard went multi-product: agro and automation each get their own view under one account. The first program is out — AXM Privacy, Windows telemetry control.

Where we work

Russia

We work across the whole country. Our test site and workshop are outside Moscow — come and see them for yourself.

Export

Deliveries outside Russia are possible — we discuss those case by case, by crop and by task.

What is happening on the farm

We run the «No Harvest Farm» channel on Telegram: the field, the machines, what broke and what worked.

Come out to the farm — we will show you everything in person

Get in touch