Clinical AI
✨ Deutschlandstipendium Scholar (2025–2026)
Available for Working Student Roles

I build practical AI tools that doctors actually use.

Clinical AI researcher at UMG Göttingen engineering evidence-grounded RAG assistants and explainable AI tools. Master's student in Applied Computer Science (1 of 17 seats) at the University of Göttingen.

See What I Build → Hire Me: Working Student ML
UMG Göttingen Research Assistant (Machine Learning)
Dual-Evidence RAG Nephrology Assistant (KDIGO 2026 + PubMed)
1 of 17 Seats M.Sc. Applied Computer Science, Göttingen
5,000+ Community Founder of nationwide student guidance network
Technical Focus

What I Build in Clinical AI

Clinical RAG

Evidence-Grounded RAG

  • Dual-source retrieval: Distinguishes clinical practice guidelines from research publications.
  • Intact clinical tables: Extracts complex staging grids via pdfplumber Markdown formatting.
  • True page citations: Maps logical page numbers matching physical printed pages via PyMuPDF.
Explainability

Explainable AI (XAI)

  • Visual saliency heatmaps: Grad-CAM integration for transparent, auditable model inspection.
  • Multi-model ensembles: Fusion architectures (ResNet, EfficientNet, DenseNet) for high reliability.
  • No blind black boxes: Designed so clinicians and analysts can verify why outputs are generated.
Foundations

Engineering & Mathematical Depth

  • First-principles math: Deep background in Linear Algebra and Calculus (Chegg Advanced Math SME).
  • Relational data skills: Production database querying and data modeling (Cognizant PL/SQL).
  • Modern toolchain: Python, LangChain, FAISS, FlashRank, Streamlit, Docker, and FastAPI.
Featured Work

Selected Projects

All Projects →
Clinical RAG & Decision Support KDIGO 2026 · PubMed

Nephrology Evidence Assistant

Clinical decision support querying practice guidelines (KDIGO 2026 AKI/AKD) and peer-reviewed literature with preserved clinical staging tables and true printed page grounding.

LangChain FAISS FlashRank Qwen-30B Streamlit
Active RAG Dev & Eval Read Architecture →
Computer Vision & XAI FastAPI & Docker

Explainable Forensic Detection

Multi-architecture fusion classifier paired with Grad-CAM visual heatmaps, giving auditors transparent spatial explanations rather than black-box decisions.

PyTorch Grad-CAM FastAPI Docker Streamlit
99.97% Celeb-DF v2 View Tech Stack →

Looking for a Clinical ML Working Student?

Available up to 20h/week in Göttingen, hybrid, or remote across Germany.

Start a Conversation →