Skip to content

Data Engineer · Deinze, Belgium

I automate work that used to be done by hand.

Geospatial production for aerial survey — computer vision on sonar and imagery, pipelines that run unattended, and the quality control that happens before anything reaches a client.

See the workGet in touch
Danil Zhuravlov, Data Engineer
Geospatial & Machine Learning
6
Production tools shipped
2 of 5
Programmers on the data team
~2 h
Saved per colleague, weekly
1000s
Detections reviewed by hand

Selected work

Computer vision · Sonar · 2025

Seabed object detection

A computer vision pipeline that finds boulders on the seabed from multibeam and side-scan sonar. Built from scratch, now in production.

  • Data preparation and point cloud to image conversion, so sonar returns could be handed to a vision model at all.
  • Model fine-tuning and evaluation, including where false positives and negatives cluster.
  • The QC workflow the survey team uses day to day — operators now review model output instead of marking every object by hand.
  • Python
  • PDAL
  • Point clouds
  • Fine-tuning
  • QGIS

Also shipped

Six tools that removed manual steps→2025

Automation · Internal tooling

Integrity checks on generated imagery, blurring of sensitive areas to meet publication rules, and convention-based file handling.

  • Given simple interfaces and made portable, so non-technical colleagues can run them anywhere.
  • Some save a colleague around two hours a week; others replaced steps that needed constant attention and now run unattended.
  • Python
  • pandas
  • GDAL
  • Packaging
Quality control before delivery→2025

Machine learning · QC

Checking machine learning output before it reaches clients — correcting detections and analysing where the model fails.

  • On one project the review covered several thousand trees.
  • False positive and negative clustering fed back into the next round of training.
  • Evaluation
  • QGIS
  • GeoTIFF
Asbestos monitoring & LOD2 delivery→2025

Geospatial production

Running a client's asbestos monitoring pipeline on aerial imagery, and coordinating LOD2 batches with an external production partner.

  • Around two LOD2 batches a month, checked and handed on.
  • Coordinate systems, LAS/LAZ and GeoTIFF handling throughout.
  • LOD2
  • LAS/LAZ
  • Coordinate systems

About

Data engineer working on geospatial production for aerial survey. Most of my job is automation and machine learning applied to work that used to be done by hand.

One of two programmers in a five-person data team, at a company where most staff are non-technical. That shapes how I build: tools get simple interfaces and go where the work is, because the people running them are surveyors, not developers. I also do the quality control on deliverables before they reach clients.

At Vansteelandt BV since March 2025.

Programming
Python · pandas · NumPy · SQL · Git · AI coding assistants
Geospatial
QGIS · GDAL · PDAL · LAS/LAZ · GeoTIFF · Coordinate systems · LOD2
Machine learning
Data preparation · Fine-tuning · Evaluation · False positive analysis
Web
Webflow · HTML · CSS · Astro

Training

  • AI & Data Science BootcampBeCode2024
  • Dutch A1–B1 certificationCVO Groeipunt2022 – 2023

Languages

  • UkrainianNative
  • EnglishProfessional
  • DutchB1

Contact

Open to work on data, geospatial and machine learning.