2025 — Present
Working Student, ML Engineer · Charité
Development of a PyTorch pipeline for clinical respiratory-audio classification, including preprocessing, cross-validation, architecture comparison and reproducible HPC workflows.
About
I am a computer-science master's student and machine-learning engineer based in Berlin. My work sits between applied ML, software engineering and human–computer interaction.
I am drawn to projects that combine technical depth with a clear real-world purpose. At Charité, I currently work on respiratory-audio classification. Alongside this, I develop open-source tools and full-stack AI applications.
Education
- M.Sc. Computer Science · Freie Universität Berlin · 2024 — Present
- B.Sc. Human-Computer Interaction · University of Hamburg · 2021 — 2024
Languages
- German, Italian & English · Fluent
- Spanish · C1
- French · B2
Experience
2024 — 2025
Research Assistant, Machine Learning · University of Hamburg
Work on unsupervised and semi-supervised anomaly detection, together with a frontend interface and Python ML backend for industrial inspection.
2016 — 2025
Marketing & AI Automation · Italian–German consultancy
Bilingual marketing (German & Italian) for an economic consultancy working across both markets, later extended with internal graphics tooling and an AI-assisted content-production workflow.
Selected projects
Open-source developer tool
WizardFlow
A small Python SDK and client-side viewer for recording, replaying and inspecting AI-agent runs as messages moving through a graph. Traces are plain JSONL files; no account or server is required.
Full-stack AI application
cs-modulio
An AI study consultant for computer-science students at Freie Universität Berlin. It answers questions from local documents, presents course information and checks proposed study plans against deterministic degree rules. Provisioned the production infrastructure on Azure using Terraform, while retaining Docker Compose for local integration testing.
Advisory only; official university documents remain authoritative.
GitHub
Independent continuation
Emergency-care dashboard
A doctor-facing demonstration for reviewing structured intake data, AI-assisted triage suggestions and subsequent clinician decisions. It is not a deployed medical product or medical device.
Built as an independent continuation of a university group project on patient intake.
Applied ML research
Respiratory audio classification
An end-to-end research pipeline for classifying short respiratory voice recordings. It compares custom architectures with pretrained audio feature extractors using grouped data splits, Optuna tuning, repeated-seed evaluation and automated result reporting.
GitHub
