janzimny@yahoo.de

Hi, my name is

Jan Zimny

Software Engineer and Machine Learning Enthusiast

I'm a software engineer specializing in machine learning and NLP, with a background in research. My work on automated detection of misleading visualizations was published at ACL 2026. Currently at SAP SE, building data-driven tools and developing enterprise software used by 6,000+ customers worldwide.

See my work Get in touch

01. About Me

Hello! I'm Jan, a software engineer and applied ML researcher. I enjoy working at the intersection of software engineering and machine learning. From fine-tuning vision-language models on GPU clusters to shipping production backend systems at enterprise scale.

I completed my M.Sc. in Computer Science in May 2025 at TU Darmstadt with distinction. My thesis on automated detection of misleading data visualizations, conducted in collaboration with KU Leuven (Belgium), was awarded the best possible grade and led to a paper accepted at ACL 2026.

At SAP SE I've built full-stack enterprise HR software, and initiated and led a weekly AI knowledge-sharing series for my team.

Here are some technologies I've been working with recently:

Enterprise / Backend

  • Java
  • Javascript
  • Spring Framework
  • SAP CAP
  • SAPUI5
  • SAP Fiori
  • SQL
  • Docker
  • Kafka
  • Kubernetes
  • Git
  • GitHub Actions (CI/CD)
  • Linux

ML / Research

  • Python
  • PyTorch
  • HuggingFace
  • MCP
  • LangChain
  • ChromaDB
  • Pandas
  • Accelerate
  • Slurm
  • A2A Protocol
  • Jupyter Notebook
  • Matplotlib

Tools & Platforms

  • SAP BTP
  • SAP SuccessFactors
  • Splunk
  • Jira
Jan Zimny

02. Experience

Software Engineer @ SAP SE

Oct 2019 – Present · St. Leon-Rot, Germany
Part-time (0.4 FTE) during M.Sc., Oct 2021 – Apr 2025

  • Full-Stack Development: Backend-focused development in an 8-person agile team for the SAP SuccessFactors Time-Off module, used by 6,000+ enterprise customers worldwide.
  • AI Knowledge Transfer: Initiated and lead a weekly AI competency workshop series on leveraging AI in software development. Authored team-wide end-to-end guidelines for AI-assisted development workflows.
  • Performance Optimization: Served as team Performance Contact — identifying and resolving critical bottlenecks in the Time-Off module through profiling and targeted fixes.
  • Production Reliability: Independently developed patches for critical customer issues; on-call duty and triage coordination with first-level support.

Department Rotations @ Fresenius SE

Aug 2016 – Aug 2019 · Bad Homburg, Germany
Dual-degree program alongside B.Sc. at DHBW Mannheim

  • IT Security (Nov 2018 – Aug 2019): Automated security-incident detection and response workflows using Splunk and Splunk SOAR, aligned with BSI and NIST frameworks. Delivered IT security awareness workshops for executives across multiple departments.
  • Mobile Development & PM (Jan 2017 – Aug 2018): Requirements analysis, design, and implementation of mobile apps including a SAP Fiori service-request application and a native Android app. Coordinated project management for an international SAP ERP rollout.

03. Some Things I've Built

Master's Thesis and Resulting Publication · 2025/2026

Is This Chart Lying to Me?

Automated detection of misleading data visualizations. My master's thesis research in collaboration with KU Leuven (Belgium) and TU Darmstadt. Helped derive the annotation schema for the real-world misleading visualization dataset Misviz, which includes 2,604 real-world misleading charts, and built a data generation pipeline to create a synthetic dataset Misviz-synth with 57,665 synthetic misleading data visualizations. Fine-tuned classifiers and generative vision-language models for chart-deception detection. Accepted at ACL 2026, San Diego. Created in equal contribution with my supervisor, Jonathan Tonglet.
  • Python
  • PyTorch
  • HuggingFace
  • VQA
  • Slurm / HPC
Paper Master Thesis Github Repo
Is This Chart Lying to Me?

University Project · 2024

Smart Reply Generation via RAG

RAG-based message retrieval system for generating contextually enriched smart replies in a job matching application, using context from previous chats. Similarity matching is done via cosine similarity of vector representations, with no custom model trained. A seminar study with participants assuming job-matching personas found high conversational relevance, perceived naturalness, and strong likelihood of adoption.
  • Python
  • RAG
  • Vector Embeddings
  • Cosine Similarity
  • NLP
GitHub Seminar Thesis
Smart Reply Generation via RAG

Bachelor's Thesis at Fresenius SE · 2019

Network Intrusion Defense System

Selection, development, and testing of an automated reactive defense system against internal network attacks within a global enterprise network. Evaluated end-to-end against the WannaCry ransomware case study, demonstrating automated containment and incident-response capabilities. As-Is and To-Be state and process documented via BPMN. Aligned with BSI and NIST frameworks. Graded 1.4 (German scale equivalent to an A).
  • Python
  • Splunk
  • Splunk SOAR
  • Network Security
  • NIST / BSI
Smart Reply Generation via RAG

Other Noteworthy Work

a handful of things I've shipped

2026

What Do Travellers Really Think About Swiss Train Stations?

A small side project analysing 22,000+ Google Maps reviews across 61 SBB stations. Each review was passed through a keyword filter and an aspect-based classifier (GPT-4o-mini) to tag which dimension of the station was being discussed and the sentiment. Output is a ranked list of action items per station, split into damage control and visible wins.

PythonPandasGPT-4o-miniWeb ScrapingNLP
Live Analysis GitHub
2024

Foodbank

University course project. A mobile app concept for reducing food waste. Keep track of items in your fridge that are about to expire and share surplus food with people nearby.

FigmaUX DesignPrototyping
Figma Prototype
2023

Geo-Diverse Common Inference Model: GD-COMET

Course project for NLP-Commonsense during my semester abroad at the University of British Columbia. Fine-tuning of COMET, a commonsense knowledge graph embedding model, on a geo-diverse dataset to improve culturally specific commonsense reasoning. One group member later developed the work into a published paper. I am listed in the Acknowledgements.

PythonPyTorchNLPSlurm
arXiv Paper
2018

Development of a SAP Fiori Mobile Application for Viewing SAP Service Notifications

Developed a SAP Fiori mobile application for viewing SAP system service notifications at Fresenius SE on the go. The app provided a user-friendly interface for service technicians to access, filter and manage their service notifications on the go.

SAP FioriJavascriptMobile
2017

Android Native Application for Viewing .tiff file formats

Developed a native Android application for viewing .tiff file formats at Fresenius SE. The app supports multi-page TIFFs, zooming, and panning, providing a user-friendly interface for professionals who need to access and analyze TIFF images on the go. The app is intent registered to automatically be recommended for opening .tiff files.

JavaAndroidMobile

04. Publications

ACL 2026 · San Diego · Forthcoming

Is this chart lying to me? Automating the detection of misleading visualizations

Jonathan Tonglet*, Jan Zimny*, Tinne Tuytelaars, Iryna Gurevych (2025)

* equal contribution

My contribution Developed the Misviz annotation schema; built the Misviz dataset (2,604 real-world misleading visualizations) and Misviz-synth (57,665 synthetic visualizations); fine-tuned classifiers and generative vision-language models for automated misleading-chart detection.
Paper GitHub

What's Next?

Get In Touch

I'm currently open to new opportunities. Whether you have an interesting project in mind, want to talk applied ML or engineering, or just want to say hi. My inbox is always open!

Say Hello