● Applied AI & robotics consulting

AI and robots that work in the operation you already run.

We design, build and run AI agents and robotic systems inside real plants, sites and fleets: connected to the systems you depend on, measured in production, secure from day one.

AI-generated footage

The gap

The model is rarely the problem. The last mile is.

37%

of organisations attribute any profit impact to AI.

McKinsey, State of AI, 2026 ↗
40%+

of agentic AI projects will be cancelled by the end of 2027, Gartner predicts.

Gartner, June 2025 ↗
5M

industrial robots now work in factories worldwide, a record.

IFR, World Robotics 2026 ↗
1.9M

US manufacturing jobs could go unfilled through 2033.

Deloitte & The Manufacturing Institute ↗

Robots and AI are ready. Making them work with your ERP, your plant network, your people and your safety rules is where projects stall. That is the work we do.

Software and robots, one team

We map the whole system before anything moves.

The agent in your dispatch software and the robot on your floor are one problem. We treat them as one.

Approach

From a real problem to a system that runs.

01 · Find

Find the work worth automating.

We go to the floor, the site and the back office, and find where AI or a robot pays back, and where it doesn't.

Output: a short list, with the numbers

02 · Build

Pilot in production, not a sandbox.

We build inside your real systems and measure against your baseline, with your team beside us.

Output: a working system, measured

03 · Run

Run it, then hand it over.

Monitoring, incident response and documentation, until your team owns it or we keep running it for you.

Output: a system your people trust

AI-generated footage

Industries

Built for operations with real-world stakes.

A white robot arm at an assembly station

Manufacturing

Cells, inspection and the planning around them.

An autonomous mobile robot in a warehouse aisle

Logistics

Mobile robots, yards and the systems that dispatch them.

An engineer at a laptop beside a robot arm

Construction

Progress tracking, inspection and site documentation.

A battery storage yard with an inspection robot

Energy

Solar, storage and substations, inspected and documented.

An inspection robot in a data center electrical room

Data centers

Commissioning evidence and operations, at build speed.

A technician with a tablet at a wind turbine

Field service

Agents that schedule, brief and close the job.

Images on this page are AI-generated illustrations.

Why Leyoxa

Engineers, not slide decks.

Works with what you run.

Old software, plant networks and vendor APIs that fail in strange ways. We have shipped agents inside exactly that.

Software and robots together.

Perception, autonomy and the business systems around them, from one team.

Security from day one.

Identity and access built the way banks build them, because a robot on your network is a computer on your network.

Judged by what actually happened.

We measure systems by their real outcomes in production, not by demos or vendor dashboards.

Work

What we have built.

A dispatch desk with a headset and a tablet, trucks outside the window

Production · AI agents

Agents running live inside legacy software.

Voice and workflow agents that answer real calls, book and change appointments directly in a multi-location business's existing scheduling systems, and hand off to staff when a person is needed.

  • Live, every day, unattended
  • Tool calls into old systems, every write checked
  • Judged on completed outcomes, not vendor success flags
An inspection robot in an electrical room

Research · Robotics

Robotic inspection for electrical commissioning.

A robot, a cart or a technician's phone that checks terminations against the drawings, reads test results against limits and peers, and catches hot joints under first load.

  • Computer vision and thermal imaging
  • Engineers make every acceptance decision
  • A research area, not yet a commercial offering

Read the research notes →

A small private server on a shelf in a quiet room

Infrastructure · Private AI

A private AI platform for sensitive records.

Self-hosted models and retrieval, with personal data redacted before any model call, every request logged, and evaluations scored by checks rather than by another model.

  • Privacy gateway and egress logging
  • Agent benchmarks with provable constraints
  • Held-out test sets for retrieval quality

About

Founded by an engineer.

Leyoxa was founded by Sepehr Aflatounian, a University of Waterloo computer engineer who builds systems people rely on every day.

  • Autonomous systemsLed Watonomous, the University of Waterloo's autonomous driving team.
  • AI agents in productionVoice and workflow agents running live inside businesses' legacy software.
  • Security at bank scaleSign-in and identity for millions of banking customers.
  • Machine learningCo-founded SeedCI: cyber-attack detection and price prediction models.
  • Connected vehiclesEV charging features for Ford's connected-vehicle platform.

Start a project

Have an operation that should run itself?

Tell us what you do and where the work piles up. We will tell you honestly whether AI or a robot can take it on.

contact@leyoxa.com