Open to opportunities

Omar Orujlu

@omarorujlu

Message

I build production computer vision, OCR, and reliable LLM extraction systems.

Germany

Message

What I'm looking for

I'm looking to build production AI systems end to end, spanning computer vision, LLMs, evaluation, data pipelines, and containerised APIs in distributed teams.

I've shipped production computer vision and OCR pipelines at HARMAN International (Samsung), supporting automotive software testing used daily by QA teams in Europe and Asia.

At Technical University of Munich, I build 3D scene-understanding systems with graph neural networks, point-cloud data, and ontology-constrained generation. I own experimentation from preprocessing and Docker training to tracking, profiling, and model debugging.

I also build confidence-gated document extraction and retrieval-augmented GenAI applications, combining guided decoding, evaluation and calibration, APIs, databases, and containerised deployment.

Experience

Work history, roles, and key accomplishments

Current

Machine Learning Engineer / Research Associate

Current

Technical University of Munich

Jan 2024 - Present (2 years 8 months)

Design and train graph neural networks for 3D scene graph generation on large-scale point-cloud benchmarks. Enforce ontology constraints during generation through logit masking and own the full experimentation stack including data preprocessing, Docker training, and experiment tracking.

PyTorch 3D Scene Understanding graph neural networks OWL Docker data preprocessing