Automation · Computer vision · LLMs · Data systems
Founder & Technical Lead, Techno-Merce · MSc ETH Zürich · Microsoft Mixed Reality & AI Lab
How we can start
Offers
Many AI projects fail because nobody checked early whether the data can answer the question. This check does that first.
For: Teams with data and a concrete idea for automation, computer vision or an AI assistant.
You get:
Language models are most useful on internal knowledge, but many companies cannot send their data to public AI tools. I build assistants that run where you decide.
For: Companies with knowledge in documents, manuals, tickets or databases that has to stay private.
You get:
Once an idea has proven itself, it has to run reliably every day, on data you can trust. I design and build the system and make sure your team can own it.
For: Organisations with a validated idea or prototype in data systems, computer vision or automation.
You get:
About
One question at every scale, across twelve orders of magnitude.
Micrometres
ETH Zürich · Spherene
Machine learning that compensates thermal errors in CNC machining, automated structural simulations that run in half the time, and an MSc in robotics, systems and control.
Metres
Microsoft Mixed Reality & AI Lab
Privacy-preserving deep learning for 3D stereo matching, so camera images can be processed in the cloud without being exposed.
Hundreds of metres
ETH Zürich · Computer Vision Group
Augmented-reality localisation data for the LaMAR benchmark, which covers 45,000 m² of university buildings, courtyards and old-town city blocks, indoors and outdoors. Plus deep learning for autonomous driving on camera and lidar data.
Thousands of kilometres
Techno-Merce
Satellite intelligence for government agencies and international organisations in six countries across West and Southern Africa. Sovereignty Eye, our satellite platform, monitors land, forests and water from Sentinel-1 radar and Sentinel-2 optical imagery.
What I do
I am an engineer who turns AI into working systems. I know computer vision and machine learning from research at ETH Zürich and the Microsoft Mixed Reality & AI Lab, and I build with language models and agents, automation, and data platforms on AWS and Google Cloud.
I offer the whole path from idea to production. I check whether your data can answer your question, build the model or system, put it into operation, and hand it over to your team with documentation and training. In Switzerland I take on selected mandates in automation, computer vision, LLM and agentic systems, and data systems.
Alongside this, I am building Techno-Merce with a few partners. We turn satellite data into decisions about land, water and natural resources, and build digital platforms for public institutions, such as the WSP Digital Audit for water safety.
I am at home in international environments, from ministries in West Africa to research labs and companies in Switzerland. I work professionally in English, German, Italian and French, and Polish is my native language.
Expertise
Automation
Software and machine learning that take repetitive work off your team, from simulation and testing workflows to content and document pipelines.
Computer vision
Inspection, localisation, 3D and satellite imagery. From a quick feasibility check to a model running in production, with accuracy measured on data it has never seen.
LLM and agentic systems
Language models and agents that search, summarise and act on your own documents and systems, with your data kept under your control.
Data systems
Architecture, APIs, databases, dashboards and cloud infrastructure on AWS and Google Cloud, including field apps that work with poor connectivity.
Have a project in mind?
Projects
Computer vision
Satellite platform for government agencies and international organisations, built on Sentinel-1 radar and Sentinel-2 optical imagery. techno-merce.com
Deep-learning encoding for cloud stereo matching, with two new privacy methods based on adversarial training and channel permutation.
Augmented-reality localisation data captured for the LaMAR benchmark (ECCV 2022), covering university buildings and old-town city blocks. lamar.ethz.ch
Multi-task learning for semantic segmentation and depth estimation, and 3D object detection from lidar point clouds.
Automation
Machine-learning model that compensates thermal errors in CNC machining.
Python automation of structural validation in Ansys, cutting runtimes by 50%.
AI and software automation for cloud storage, e-learning content creation and in-house productivity, plus computer vision for image and video enhancement.
Data systems
Water Safety Plan system based on the WHO methodology: mobile field assessment, role-based access, automated scoring and a certification workflow.
Greenhouse-gas inventory model for the Ministry of Local Government, Chieftaincy and Religious Affairs and WaterAid Ghana, delivered with KNUST.
Personalised recommender, real-time analytics, AWS architecture, DevOps and data-privacy compliance.
Teaching
Computer Vision, Programming for Robotics (ROS), Introduction to Robotics and Mechatronics, and the Robotics Summer School.
Own company
Environmental intelligence for resource protection and sustainable supply chains. techno-merce.com
Recommendations
“He is intelligent, motivated, kind, takes initiative and is a great communicator, traits which make him easy to work with, and ideal for project lead and delivery.”“He brings enthusiasm to discussions and is eager to explore innovative solutions as well as to work independently.”Ready to start?