- 🔬 Building production ML systems — NLP, Computer Vision, and multimodal AI
- 🏗️ Currently crafting data pipelines & ARR models at niologic GmbH (Germany)
- 📄 Published researcher in multimodal NLP for the construction industry (CIB W78 2024)
- 🌍 Languages: Chinese (native), German (fluent), English (fluent)
- 🎓 M.Sc. Applied CS @ Ruhr University Bochum · B.Sc. IoT Engineering @ Beijing University of Technology
gantt
dateFormat YYYY-MM
title Work Experience
section Employment
ML Researcher & CV/NLP Engineer — Jaeger Gruppe :done, jg, 2021-07, 2025-01
Senior Data Scientist & ML Engineer — niologic GmbH :active, nio, 2025-01, 2026-02
section Education
B.Sc. IoT Engineering — Beijing Univ. of Technology :done, bsc, 2016-09, 2020-07
M.Sc. Applied CS — Ruhr University Bochum :done, msc, 2020-10, 2023-03
| Project | Description |
|---|---|
| digital-poststelle | Multimodal NLP email classifier for construction — BERT+BiLSTM, 92.3% accuracy. CIB W78 2024 |
| CondiNILM | Conditioned multi-task learning framework for non-intrusive load monitoring |
| ViT-HyperSense | Systematic hyperparameter sensitivity analysis of Vision Transformers with Optuna |
| ARPL-MRI | Adversarial Reciprocal Points Learning for MRI Open Set Recognition |
| Project | Description |
|---|---|
| everything-claude-code | Complete Claude Code config collection — agents, skills, hooks, commands. Anthropic hackathon winner. |
| superset-mcp-new | Connect 50+ data stores via Apache Superset MCP server |
| metabase-mcp-server | MCP integration layer for Metabase + AI assistants |
| PageIndex | Vectorless, reasoning-based RAG — 98.7% accuracy on FinanceBench |
Optimizing Email Classification in the Construction Industry through a Multimodal NLP Approach Yu Zha, Sherief Ali, Sebastian Schumacher, Michael Schulte, Markus König CIB W78 2024, Marrakesh, Morocco — PDF
Open to collaboration on ML research, AI tooling, and data infrastructure projects.