Python package · DFKI
Repository ↗
DiversityLens
A modular toolkit for auditing demographic distributions in computer vision datasets before model training and dataset selection.
Sample audit — COCO-People subset
M.Sc. Embedded Systems Engineering candidate at FH Dortmund — building DiversityLens at DFKI to audit demographic balance across large-scale computer vision datasets.
Profile
My work sits between applied computer vision, dataset auditing, and quality review. I am strongest where AI outputs need to be tested, organized, explained, and made reproducible through Python workflows, dashboards, and careful documentation.
Experience
Coursework and project work across computer vision, machine learning, software architecture, and applied AI systems.
Reviewing high-volume content quality and policy decisions for a major social media platform, with consistent accuracy and a practical eye for ambiguous edge cases.
Developing DiversityLens — a Python toolkit for demographic analysis of robot–human interaction and computer vision datasets using OpenCV, DeepFace, RetinaFace, Pandas, Bokeh, and CLI-based workflows.
Selected Projects
A modular toolkit for auditing demographic distributions in computer vision datasets before model training and dataset selection.
A privacy-preserving retrieval pipeline with Llama 3-ChatQA, FAISS, Sentence Transformers, chunking experiments, and streaming responses on consumer hardware.
Interactive Demo
Select a dataset, configure detection settings, and run a simulated demographic audit.
Toolkit
Python, Bash, C++, MATLAB, Git, Linux, LaTeX
DiversityLens — CLI pipelines
RAG System — retrieval & chunking
OpenCV, DeepFace, RetinaFace, Pandas, NumPy, Scikit-Learn, Bokeh
DiversityLens — face detection, demographic analysis, Bokeh dashboards
RAG pipelines, FAISS, Sentence Transformers, OpenAI API, LangChain
Local RAG System — FAISS retrieval, Llama 3-ChatQA, streaming inference
FastAPI, Docker, GitHub Actions, Pytest, documentation, CLI workflows
DiversityLens — Pytest suite, Docker packaging, GitHub CI
Beyond the CV
I care deeply about building AI that is honest about what it doesn't know, fair in how it treats different people, and grounded in real-world data quality — not just benchmark numbers.
Dortmund, Germany
↻ hover to flipCurious by nature, careful by practice. I find beauty in edge cases and tend to ask "what happens when this breaks?" before "does it work?"
Outside of code — hiking, reading about cognitive science, and probably overthinking a dataset.
Target Roles
Looking for junior roles where AI quality, dataset work, evaluation, and pragmatic Python tooling matter more than buzzwords.
Contact
Open to AI evaluation, data quality & applied AI roles.