Smart Tinmen logo Smart Tinmen
Product prototyping studio · robotics · sensing · AI

Robots and sensors that keep working outside the lab.

We turn early ideas into builds you can test in the field: underwater drones, weeding robots, camera rigs and the AI that runs on them. Electronics, firmware, software and deployment from one small team.

Start a project See the work
Smart TinmenStudio spec sheet
Registered
2024 · KVK, Sint Kruis
Field hardware
Since 2015, from offshore diving to ROVs
Deployed in
Greenland · Malawi
Pose accuracy
~0.2° per frame, 6DoF tracking
Stack
ROS2 · OpenCV · PyTorch · Blender

Work

Projects we have built and tested. Several are open source, so you can read the code before you call us.

Client project · underwater

Predictive maintenance for underwater drones

Models that read drone telemetry and flag parts that are wearing out, so maintenance happens on deck instead of after a failure at depth.

Telemetry · ML · ROV
Client project · agriculture

Automatic weeding robots

Perception and control for robots that find weeds between crop rows and remove them without herbicide.

Computer vision · control
Studio build · 3D capture

Automatic photogrammetry turntable

A turntable that rotates the object and triggers the camera at each step, then feeds our own pipeline from photos to a 3D model and a Gaussian splat.

Photogrammetry · Gaussian splattingcode ↗
Studio build · computer vision

6DoF tracking of tabletop miniatures

Printed ArUco markers on a hexagonal base, ChArUco camera calibration and one pose solve over every visible marker corner, to animate a 3D model from a real figurine.

~0.2° error per framecode ↗
Lilongwe Wildlife Centre · Malawi

Pangolin detection from the air

Computer vision for an autonomous drone that finds and tracks pangolins, one of the most trafficked animals. We ran the drones, wrote the flight plans, trained on Blender-generated scenes and shipped a Streamlit app for the team.

YOLO · drones · synthetic datacode ↗
Studio research · simulation

nexus population simulator

A grid-based coalescent simulator that reproduces population-stratification bias in genetic studies, used to test correction methods before they meet real data.

Simulation · statisticscode ↗

What we do

Small, hands-on engagements. You work directly with the engineers who build the thing.

Prototype builds

Electronics, firmware and software together, up to a demo that survives real conditions.

Arduino · Raspberry Pi · BeagleBone

Robotics & perception

Control, sensor integration, camera calibration, pose estimation and 3D capture.

ROS2 · OpenCV · photogrammetry

Field & wildlife monitoring

Drones, flight planning and species detection, built for remote sites with little power or bandwidth.

Drones · YOLO · Blender sim

Applied AI

Models and LLM agents wired into the product, with evaluation so you know how well they work.

PyTorch · LLMs · RAG · eval

How a project runs

Weekly demos from the first week. Decisions are written down as we go.

  1. Step 1

    Scope

    Constraints, success criteria and deliverables, agreed in a short kickoff.

  2. Step 2

    Build

    Fast iterations across hardware and software, with a demo every week.

  3. Step 3

    Test

    In real conditions: power, sensing quality, failure modes, not only on the bench.

  4. Step 4

    Handover

    Working demo, documentation and a roadmap to pilot or production.

Team

Two industrial engineers from ECAM Brussels. One runs the projects, one builds them.

Harold Loop

Harold Loop

CEO

Electromechanical engineer who delivers digitalisation projects in power plants, robotics centres and industrial sites, as product owner, analyst and on-site lead.

  • 2022–nowConsultant, energy sector · Avertim
  • 2021–2022Data engineer · Deloitte Consulting
  • 2020Thesis, marine robotics · VLIZ
Matthieu de Hemptinne

Matthieu de Hemptinne

CTO

Robotics and AI engineer. Went from commercial diving to ROV control, drone vision in Malawi, deep learning for genetics and LLM systems in production. Two M.Sc. degrees in industrial engineering.

  • 2025–2026Drone CV · Lilongwe Wildlife Centre
  • 2021–2025ML engineer · CTG, Nature Neuroscience
  • 2019–2020Marine robotics · VLIZ, Greenland

Tell us what you are building

  1. What you want to build or test
  2. Your constraints: budget, timeline, environment
  3. What success looks like
Email
info@smarttinmen.com We reply within 1–2 business days with a scoped plan and an estimate.