- 🏭 10 years production AI/ML at Eurailscout — deep learning systems for national-scale railway inspection across optical, ultrasound & eddy-current sensors
- 🤖 Now building agentic LLM systems and cloud-native data pipelines that bridge foundation models with real-world sensor and business data
- 🎓 M.Sc. Computational Science & Engineering, TU Munich (Grade 1.8) · Published researcher (ECNDT 2018 first author · WCNDT 2016)
Skills: Python · C++ · PyTorch · CUDA · TensorRT · LLM · RAG · multi-agent systems · FastAPI · AWS · Terraform · Docker · PostgreSQL · MongoDB · RabbitMQ · OpenCV · MLflow · data pipelines · computer vision · HPC
💼 Open to AI Engineer and Data Engineer roles in Berlin
LLM / Agents
Cloud
Data
ML / Vision
Languages
| Project | Description | Stack |
|---|---|---|
| iHunter | Agentic LLM platform for B2B lead discovery — multi-agent orchestration, automated data enrichment & quality assessment | Python · Groq · LLaMA 3 · FastAPI · aiosqlite |
| landlord_system | Property management for German landlords — rent tracking, utility billing, and PDF generation | Python · PostgreSQL · FastAPI |
| AWS Pedestrian Detection | Event-driven ML pipeline: S3 → SNS → Lambda → RDS · Terraform IaC | AWS · Python · HCL |
| MonoSLAMAutoCalibration | M.Sc. thesis — monocular SLAM for automatic camera self-calibration of surgical endoscopes (Gaussian Sum Filter) | C++ · OpenCV · Eigen |
| kleinanzeigen-scraper | Full-stack scraper with browser UI, price & location analytics | Python · FastAPI · PostgreSQL |
- S. Gao et al. — Multi-modal sensor fusion for automated railway defect classification · ECNDT 2018 · First author
- S. Gao et al. — Cross-sensor data fusion for rail surface analysis · WCNDT 2016
🇬🇧 English — C2 · Cambridge BEC Higher | 🇩🇪 German — C1 | 🇨🇳 Chinese (Mandarin) — Native
