About me
I am an Applied Machine Learning Engineer with 6+ years across AI research, ML development, and applied engineering. I combine mathematical reasoning, hands-on implementation, and research-driven investigation to solve challenging technical problems.
I completed postdoctoral research in Machine Learning & AI and hold a Ph.D. in Mathematics. My mathematical training taught me to break complexity into manageable parts, test assumptions, and reason clearly about possible solutions.
I apply this foundation to practical engineering: selecting and combining AI techniques, building maintainable workflows, validating behavior under imperfect inputs, and balancing quality, reliability, speed, and complexity.
When existing tools fall short, my research experience helps me read and interpret scientific literature, learn unfamiliar techniques independently, and turn promising ideas into experiments and implementable solutions.
In a Germany-based startup, I was the sole AI engineer responsible for the AI workflow of a production-facing computer vision product, working with medical specialists and a backend engineer. In my independent LLM Systems Lab, I developed modular RAG and evaluation components.
I value technical ownership, critical thinking, continuous learning, responsible execution, and clear communication with people from different backgrounds.
Technical toolkit
Python, PyTorch, OpenCV, RTMPose, DensePose, SAM2, MediaPipe; pose estimation, segmentation, landmark refinement, human detection, multi-view body analysis, confidence estimates, geometry-based measurements, and quality control.
Modular Python pipelines, scikit-learn, model benchmarking and validation, failure analysis, unit and regression tests, structured logging, latency tracking, metadata capture, reproducible experimentation, configuration-driven development, Git workflows.
RAG, embeddings, semantic retrieval, retrieval metrics, ChromaDB, LangChain, Ollama, Gemini, LLM inference, Transformer fundamentals, GNNs, graph representations, knowledge graphs, PyKEEN, link prediction, predictive modeling, Optuna, Ray Tune, and RNN/LSTM fundamentals.
Linux/Ubuntu (four years of regular use), Bash, SSH, Git/GitHub collaboration, WSL2, Docker, FastAPI-related API testing, Swagger/OpenAPI, and basic SQL. FastAPI backend workflows were used in collaboration with the backend engineer; deployment ownership is not claimed.
Educational time-series forecasting with COVID-19 case-count data; independent technical study of football video analytics: player detection, multi-object tracking, ReID, tracklet identity, trajectory refinement, team assignment, jersey OCR, and ID-switch evaluation.
In a separate private project, I am extending a RAG foundation toward application APIs, Docker Compose deployment, n8n orchestration, agentic workflows, MCP, and CI/CD. These are planned or evolving capabilities, not asserted as completed professional specializations. Project scope →
Mentorship & teamwork
I have taught graduate-level topics in graph neural networks, machine learning, and optimization, and served as a supervisor and advisor to multiple M.Sc. students covering knowledge graphs, GNNs, and LLM-based AI systems. These experiences develop my ability to explain difficult ideas, structure feedback, and support junior team members with clear, constructive technical guidance.
Teaching & mentoring → · Full experience →