software engineering · Wien, AT

Software for
high-stakes
imaging.

I’m Ulrich Mayer — a software engineer who builds reliable, regulated software for medical devices and industrial systems. Image acquisition and computer vision, the platforms that grade and review what they produce, and the on-site integration that gets it running in the clinic. Day to day, senior developer at ImageBiopsy Lab, building CE-marked and FDA-cleared AI for musculoskeletal imaging.

  • 17+ years in regulated software
  • 4 countries of on-site commissioning
  • 510(k) · CE cleared & marked systems
  • Vienna based, working worldwide

Four ways a grown system leaks time — and what I put in their place.

01

Data scattered across a dozen spreadsheets

The customer sheet, the device sheet, the transmission log, the quality list — each kept by hand, none agreeing with the others. I replace them with one data model that holds the real state of every device and every case, queried instead of reconstructed from memory.

  • Data modelling
  • Spring Boot · JPA
  • SQL & NoSQL
  • Jira / CMDB
02

A pipeline that only moves when pushed

Data copy-pasted between folders by hand, scripts run one at a time, a licensed tool for one step and an unmaintained one for the next. I turn it into an unattended flow — receive, filter, analyse, visualise — on maintained open-source and in-house tools, with the licences gone.

  • Python · FastAPI
  • DICOM · DCMTK
  • Open-source stack
  • No-touch pipeline
03

No triage, no follow-up

Nothing chases the facility that stopped sending images; nothing says which device is most overdue. I put the rules in software — raise the ticket and start the follow-up automatically, and rank the queue by severity and time since last inspection.

  • Rule automation
  • Jira workflows
  • Prioritisation
  • Scheduling
04

Failures found the expensive way

A device or software change on the customer side used to surface only as a broken evaluation, after the effort was spent. I build the checks that catch it first — numeric limits, deviation grading, and a warning to the reviewer before they start.

  • Validation
  • QA limits
  • Change detection
  • Grading rules

Most systems aren’t designed. They grow.

A spreadsheet here, a script there, a manual copy‑paste to keep it all moving. My job is to understand the process that grew, understand the environment it runs in, and put the right tools around it — so the people using it get their day back. DIRPortal is that idea, worked all the way through.

Own product · deployed at AGES

DIRPortal

DICOM image-quality review portal

A platform for the technical quality assurance of mammography and ultrasound imaging systems. It ingests DICOM reference images from healthcare facilities, dispatches them to an analysis server, grades the results against configurable limits and tagging rules, and presents everything through a keyboard-first review interface.

ImportIdentifyAnalyzeGradeReview
  • Self-contained DICOM viewer — window/level, zoom, pan, auto-windowing, a histogram windowing panel and QA-deviation overlays decoded from the analysis thumbnails.
  • Three-tier automatic tagging of findings: global → manufacturer → device, with per-label overrides and a rename/merge tool.
  • Per-device numeric limits that grade findings into bands and can auto-apply tags.
  • Per-user, customizable dashboards — an add / remove / reorder card grid gated by modality role.
  1. BeforeA dozen Excel lists kept in sync by hand — customers, devices, transmissions, device quality, correspondence — plus analysis scripts scattered across MATLAB and Python.

    NowOne source of truth. All of it lives in DIRPortal and Jira.

  2. BeforeIncoming data copy-pasted between folders by analysis state, pushed through the pipeline by hand, run first through licensed third-party pre-processing and then the right analysis tool for that exact case.

    NowReceive → filter → analyse → visualise runs end to end with no user interaction, on open-source and in-house tools that are maintained — and no licence cost.

  3. BeforeFacilities that never sent their images were chased manually and tracked in a spreadsheet.

    NowRules open a Jira ticket and start the follow-up process automatically.

  4. BeforeNo way to prioritise — urgency was judged by eye, if at all.

    NowDevices ranked automatically by severity and time since last inspection.

  5. BeforeA device or software change on the customer side showed up only as a failed evaluation — then someone had to find out why.

    NowDetected automatically and flagged to the reviewer before the evaluation even starts.

Most of the effort used to go into keeping the system running. Now it goes into the quality assurance that was the point all along.

Also in flight

Own product · in development

Klopago

A bilingual (German / Croatian) app for pre-ordering school lunches. Restaurants publish their weekday menus — photo, ingredients, vegetarian flag, Austrian A–R allergen codes — and release a week for ordering; parents manage several children, set a standing preference and opt each child in or out day by day, with a monthly consumption overview.

  • Menu lifecycle with approval gates: draft → released → retired, and a per-week release that controls what parents can see at all.
  • Allergen model on both sides — menus and children each carry an allergen set, so conflicts surface before the order.
  • Scheduled reminders for undecided days and a daily per-menu totals report for the kitchen; GDPR erasure built in from the start.
  • Spring Boot 4 · Java 25
  • Angular 22 · Material
  • MariaDB
  • JWT · RBAC
  • Docker

Own product · concept

Training-data management platform

A platform for curating the datasets that machine-learning models are trained on. Data lands in cloud object storage, gets assigned to annotators, is labelled, then passes a second-pair-of-eyes review before it is released for training — with every item’s state, owner and reviewer tracked through the whole chain.

  • Role-based access — administrator, labeler, reviewer — enforced against the corporate directory, so nobody reviews their own work.
  • Assignment and review workflow with a full audit trail from upload to release.
  • Python · FastAPI
  • Azure Blob Storage
  • Entra ID / MSAL
  • React
  • RBAC

The tools, kept sharp.

Languages

  • C / C++
  • Java
  • Python
  • TypeScript
  • Matlab

Imaging & CV

  • VTK
  • ITK
  • RTK
  • DCMTK
  • KDL
  • CUDA
  • TensorFlow
  • pydicom

Visualisation

  • Cornerstone
  • dicom-parser
  • Chart.js
  • ng2-charts
  • jsPDF
  • Angular Material

Platforms

  • Angular
  • Spring Boot
  • FastAPI
  • Qt

Databases

  • SQL
  • NoSQL
  • MariaDB
  • PostgreSQL
  • MS SQL Server

Practice

  • ISO 13485
  • IEC 62304
  • 510(k)
  • Distributed systems
  • IPMA-D

Seventeen years next to machines that matter.

  1. Independent Software Developer

    Independent practice — Dipl.-Ing. Ulrich Mayer

    • Registered trade since March 2021: services in automatic data processing and information technology (GISA 33554358, Magistrat der Stadt Wien).
    • Custom software on contract — imaging pipelines, regulated platforms and full-stack systems. This is the practice this site is about.
  2. Senior Software Developer

    ImageBiopsy Lab GmbH

    • AI software for automated musculoskeletal X-ray and CT analysis — knee osteoarthritis, leg and hip geometry, spine and bone-age assessment — CE-marked and FDA-cleared, in use at 100+ sites across more than ten countries.
    • Deep-learning pipelines and the clinical platform around them: DICOM ingest, PACS and RIS integration, measurement and reporting, deployed into hospital environments.
  3. Software Engineer

    medPhoton GmbH

    • 510(k)-cleared software for medical devices; acquisition, visualization and registration of 2D/3D CT images.
    • On-site integration at ProNova / Provision Cares Proton Therapy (Knoxville), Accuray (Sunnyvale), Mevion / MAASTRO (Maastricht) and ProTom / Massachusetts General Hospital (Boston).
  4. Software Engineer / Research Assistant

    Paracelsus Medical University — radART

    • Software for the world’s smallest cone-beam CT medical imaging system; CE-certified under ISO 13485 and IEC 62304.
    • Treatment-workflow planning and execution software, from first idea to clinical release; on-site integration at MedAustron, Wiener Neustadt.
  5. Technical Adviser Energy / Software Engineer

    Remas System GmbH

    • Software for embedded Linux devices; real-time energy-consumption monitoring and control, with on-site hardware installation.
  6. Hardware Engineer · Trainee

    Sony DADC · Dr. Johannes Heidenhain · ESL Elektro

    • Hardware and firmware roots — including precision linear-measurement work at Heidenhain, where measuring to tolerance was the whole job.

Have something that has to be right?

Whether it’s an imaging pipeline, a regulated platform, or a system that just refuses to behave on-site — tell me about it. I take on focused, senior engagements.

ulrich@mtwo.at

Kontakt · Eckdaten

Dipl.-Ing. Ulrich Mayer

Penzinger Strasse 159/4
1140 Wien · Austria

VAT
ATU 76573806
GISA
33554358 — Magistrat der Stadt Wien
Trade
Dienstleistungen in der automatischen Datenverarbeitung und Informationstechnik