I develop computer vision systems that turn images into useful decisions, from industrial inspection to understanding retail shelves.
At Scandit, I coordinate Scene Parsing work in ShelfView and contribute with my team to the product recognition pipeline. With AI-powered coding agents now part of my workflow, my focus has shifted toward defining problems, directing implementation, and reviewing results, drawing on years of hands-on ML engineering.
Object detection & recognition2D & 3D classification and segmentationProduction ML systems
01 / SELECTED WORK
Different domains. The same engineering discipline.
Model development, practical constraints, and the path from data to production.
RETAIL / DETECTION + RECOGNITION
SCANDIT · SHELFVIEW
Understanding products on retail shelves
I contribute to the development of the Product Recognitino pipeline for ShelfView, coordinating the detection pipeline and enabling recognizing products in shelf modules.
The work connects computer vision model development with the requirements of a production retail application.
Object detectionRecognitionProduction ML
More about the work +
My focus is building a robust object detector within Scene Parsing. The challenge spans identifying individual products in shelf imagery and supporting product recognition within the broader ShelfView system.
I work with product and data teams to connect model development and evaluation to the application’s needs.
INDUSTRIAL CT / VOLUMETRIC SEGMENTATION
MICROTEC · INDUSTRIAL INSPECTION
Computer vision inside the log
At Microtec, I worked on 3D voxel segmentation of CT scans of logs, applying deep learning to volumetric industrial imagery.
This work extended my computer vision experience beyond surface images into three-dimensional data used for wood inspection.
3D segmentationCT imagingDeep learning
More about the work +
The central task was voxel-level segmentation in CT volumes. It formed part of my broader work developing neural networks for industrial scanners at Microtec.
Across that work, I contributed to the process from data preparation and augmentation to architecture selection and production deployment.
SURFACE INSPECTION / DEFECT GRADING
MICROTEC / BIOMETIC · QUALITY INSPECTION
Grading quality from external defects
I developed classification and segmentation models for grading fruit and boards based on external defects.
Networks I trained were deployed in production environments from early 2018, across fruit, board, and log applications.
ClassificationSegmentationIndustrial deployment
More about the work +
I collaborated with operational teams on data collection and worked through cleaning, augmentation, and architecture selection, balancing accuracy requirements with execution-time constraints.
These applications connected model outputs to practical inspection and grading tasks across Microtec and Biometic.
02 / EXPERIENCE
Models are one part. The system matters too.
Hands-on ML development alongside ownership of the tools and processes that support it that enable everyone in the company to generate meaningful impact.
2022 — PRESENT
Scandit
Senior Machine Learning Engineer
Coordinating Scene Parsing work and contributing to the product recognition pipeline for ShelfView. Since adopting Claude at Scandit, my day-to-day work has shifted toward directing coding agents: framing tasks, guiding technical decisions, and reviewing their output. I bring hands-on ML experience to that process and remain accountable for the quality of the resulting work.
2017 — 2022
Microtec / Biometic
Deep Learning Specialist
Developed neural networks for industrial inspection, from data preparation through deployment.
Training framework leadership, 2020–2022 +
As project lead, I developed the shared company framework for training neural networks. It standardized internal data formats, architecture development, training workflows, and model deployment into scanners.
The framework supported classification, segmentation, and regression, with local or training-server execution using Docker and Jenkins. It was used across Microtec and Biometic.
2018
Hawk-Eye Innovations
Freelance Computer Vision Engineer
A consulting engagement alongside my Microtec role, focused on segmentation and video matting.
2016 — 2017
Previnet
R&D Data Scientist
Healthcare-focused analytics and software engineering, including real-time data processing and medical text analysis.
03 / EXPERTISE
From training data to deployment.
01
Computer vision
Detection, recognition, classification, and semantic segmentation in 2D images and 3D volumes.
PyTorch · TensorFlow · Keras
02
Production engineering
Reusable training workflows, data standardization, model evaluation, and deployment integration.
Python · Docker · CI/CD
03
AI-assisted technical leadership
Translating engineering goals into well-defined tasks for coding agents, directing their work, and applying technical judgment to review and refine the results.
Claude · Agent orchestration · Technical review
04 / RESEARCH & EDUCATION
A foundation in image analysis.
My PhD research explored computer-aided analysis of confocal endomicroscopy images, using engineered features and random forests for segmentation and classification.