AI systems / simulation / data quality
M.Sc. Embedded Systems Engineering candidate at FH Dortmund. I build practical AI and simulation systems with a focus on dataset auditing, edge deployment, evidence-grounded retrieval, and source-backed biological modeling.
Computational biology, stochastic simulation, computer vision dataset auditing, and embedded AI.
Build the pipeline, document the assumptions, test the outputs, and keep the evidence trace visible.
Junior roles in AI evaluation, research engineering, data quality, and applied AI tooling.
Experience
A compact view of the work that shaped the portfolio: thesis research, high-volume quality review, and engineering projects that make model behavior inspectable.
Completed DiversityLens, a Python toolkit for demographic analysis of robot-human interaction and computer vision datasets using OpenCV, RetinaFace, DeepFace, Pandas, Bokeh, pytest, CI, and CLI workflows.
High-volume content quality and policy review work for a major social media platform, with attention to ambiguous edge cases and consistent decision quality.
Coursework and project work in machine learning, computer vision, model-based systems engineering, microelectronics, and hardware/software co-design.
Selected projects
Each project has a direct demo or repository link. The demos are deployed inside this portfolio so reviewers can inspect the work without hunting through local folders.
Dataset-auditing toolkit for demographic analysis across large-scale computer vision datasets. The work scaled to roughly two million analyzed images and produced reproducible reports for bias and balance inspection.
A compact neural-net project that trains an MLP in PyTorch with a manual training loop, exports weights and biases, and moves toward C++ int8 inference for edge deployment.
A source-grounded stochastic whole-cell simulation of a hepatocyte with a Python biochemical engine and TypeScript/Three.js visualizer. The project tracks measured targets and keeps assumptions explicit.
A semantic document search and grounded chat system. It ingests TXT, PDF, and DOCX files, chunks text with overlap, embeds with sentence-transformers, and retrieves cited passages by cosine similarity.
Toolkit
A practical stack for AI systems that need more than a demo: data processing, evaluation, reproducibility, and a clear path from model output to evidence.
Python, Bash, TypeScript, C++ fundamentals, Git, Linux, LaTeX.
PyTorch, scikit-learn, OpenCV, DeepFace, RetinaFace, NumPy, Pandas.
RAG workflows, FAISS, Sentence Transformers, Hugging Face, FastAPI, SQLite.
Gillespie SSA, chemical Langevin, reaction-diffusion, numerical methods, Three.js, pytest.
Contact