Mohammadreza GhassemzadehComputer Scientist / ML Engineer / AI Researcher
I build machinelearning systems and investigate how they learn, where they fail, and whether their decisions can be trusted.

Working across machine learning, applied AI, computer vision, and data science.
Work
- 01Flagship
Brain MRI Classification & Explainable AI
Does a high score on public brain MRI collections show what a model learned about tumors, or what it learned about the datasets?
A study of whether brain tumor classifiers learn meaningful features or exploit shortcuts in their training data, using cross-dataset evaluation and saliency methods.
M.Sc. thesis · University of Hamburg & SDSU
PyTorch · CNNs · Vision Transformers · Grad-CAM · HiResCAM
- 02
Computer Vision for Traffic Safety
The detection and tracking pipeline for a Caltrans-funded intersection safety study: from aerial drone footage to vehicle trajectories and safety measures.
SDSU STAR Lab · Graduate Research Assistant
YOLO · Multi-object tracking · Video pipelines
- 03
Data Engineering at Otto Group one.O
Nearly three years of SQL pipelines on Oracle, validation, automation, and monitoring behind the reporting that business teams rely on.
Data & Automation Engineer · Working Student
Oracle · SQL · Python · Power BI · Elastic Stack
- 04
Applied AI & Large Language Models
Fine-tuning, retrieval, voice agents, and an LLM application built for a client, each measured against the task it was meant to do.
Five projects · 2024 – 2026
LoRA · RAG · OpenAI API · SentenceTransformers

I like to know whya number is what it is.
I studied computer science at the University of Hamburg, first the bachelor’s and now a master’s with a focus on data science. Since January 2026 I have been a visiting graduate student at San Diego State University, where I work in the STAR Lab and write my thesis.
Alongside my studies I spent four years in production engineering: Java and .NET applications at Prinovis, then data and automation work at Otto Group one.O. That is where I learned that a wrong figure in a dashboard usually has a cause several steps upstream, and that the fix belongs there.
The same habit carries into research. When a model scores well, I want to know what it is using to get there. My thesis and most of my recent projects come back to evaluation: how a result was measured, and whether it would hold on data the model has not seen.
- Languages
- English (fluent) · German (native) · Farsi (native) · Spanish (conversational)
- Honors
- SDSU Global Campus Dean’s List, Spring 2026
Experience
- Jun 2026 – PresentSDSU STAR LabGraduate Research AssistantDetection and tracking pipeline for a Caltrans-funded intersection safety study, built on aerial drone footage.
- Feb 2023 – Dec 2025Otto Group one.OData & Automation Engineer, Working StudentSQL views and scheduled jobs on Oracle behind recurring reporting, plus validation, automation, and monitoring.
- Sep 2021 – Jan 2023Prinovis (Bertelsmann)Software Developer, Part-TimeMigrated production applications from Java 8 to 17 and .NET 4.5 to 4.8, and maintained the Gradle build and Jenkins CI pipelines.
Education
- Oct 2024 – Dec 2026University of HamburgM.Sc. Computer Science (Data Science)Expected December 2026. Thesis on brain tumor MRI classification with explainable AI.
- Jan 2026 – Dec 2026San Diego State UniversityVisiting Graduate StudentGPA 3.9 / 4.0. SDSU Global Campus Dean's List, Spring 2026.
- Oct 2020 – Sep 2024University of HamburgB.Sc. Computer ScienceThesis with Otto Group one.O on a framework for matching heterogeneous product datasets.
Tools I work with.
Languages
Python, SQL, Java, Bash, C and C++ (project experience), C# and Rust (basic)
Machine Learning
PyTorch, Hugging Face Transformers, scikit-learn, CNNs, Vision Transformers, LoRA fine-tuning, Grad-CAM, HiResCAM
Computer Vision
YOLO, Object detection, Multi-object tracking, OpenCV, Video pipelines
LLM Systems
OpenAI API, LangChain, RAG, Embeddings and retrieval, Agentic workflows, Claude Code
Evaluation
Metric design, Ablations, Leakage detection, Benchmark protocols, Error analysis
Data Engineering
Oracle, PostgreSQL, SQL query optimization, ETL pipelines, Scheduled jobs, Apache Spark, Data validation, Power BI
Engineering & Platforms
Git, Jenkins CI, Gradle, Unit testing, REST APIs, Linux, GCP, Azure, Elastic Stack
QuestionsI keep asking.
- Trustworthy AIWhen can a model's output be relied on?
- Explainable AIDoes an explanation show what the model actually used?
- Medical Computer VisionWhat does a classifier see in a scan?
- Model GeneralizationDoes performance hold on data from somewhere else?
- Shortcut LearningDid the model learn the task, or a pattern that happens to come with it?
- Data QualityHow much of a result is decided before training starts?
- Robust Machine LearningWhich changes in the input should leave the answer unchanged?
- Applied AIWhat does it take to make a model useful inside a real system?
Beyondthe code.
Travel, field days, and ordinary moments away from the screen. Nothing here is measured or classified.
Have an interestingproblem to solve?
I am happy to talk about machine learning, evaluation, computer vision, and data engineering roles, and about research collaborations.