AI & Machine Learning Engineer

Building agentic AI systems that make it to production.

AI/ML engineer with 5+ years of Python-based development across generative AI, computer vision, NLP, and embedded ML. Currently building a production multi-agent GenAI system (LangChain, LangGraph, RAG, FastAPI) at ARRK Engineering for a large-scale BMW initiative — with a research background spanning diffusion models, computer vision, and TinyML. M.Sc. in Artificial Intelligence (FAU, grade 1.3).

View CV Get in touch

Toolbox

Skills

Technologies and tools I work with across the AI/ML stack.

Agent, LLM & NLP

LangChainLangGraphLangFuseMulti-agent systemsAgentOps/MLOpsRAG pipelinesPrompt & context engineeringWeaviate (vector DB)

AI & Machine Learning

PyTorchPyTorch LightningTensorFlowKerasHugging Facescikit-learnOpenCVWorld ModelsLatent Space Modeling

Infrastructure & DevOps

FastAPIDockerDocker ComposeKubernetesHelmCeleryGitHub ActionsPostgreSQLAWS (CDK)CI/CD

Languages

PythonCC++SQLGitLaTeX

Career

Experience

December 2025 – Present

Munich, Germany

AI Engineer · ARRK Engineering GmbH

Design and develop a production multi-agent Generative AI system (orchestrator plus specialized subagents) serving 10+ active users on a large-scale BMW initiative, using LangChain, LangGraph, and LangFuse for orchestration, tracing, and evaluation. Built a RAG pipeline with Weaviate and PostgreSQL, scalable FastAPI backends containerized via Docker, and Kubernetes/Helm deployments with GitHub Actions CI/CD. Core contributor on a 5-person team, running stakeholder demos and serving as go-to resource for agent design and LLM integration — alongside exploratory research on world models for driving scenarios.

October 2023 – November 2025

Tennenlohe, Bavaria

Working Student – AI & Machine Learning · Fraunhofer IIS

Built and optimized computer vision pipelines across two applied research projects (SyNaKI, GAIA) using TensorFlow/TFLite, model compression and quantization, and custom OpenCV algorithms — deploying models on embedded hardware. Implemented a distributed AI tool using evolutionary algorithms and supported AI research funding applications.

March 2022 – March 2023

Deggendorf, Bavaria

Research Assistant · DIT Lab for Digitalization of AI in Electrical Engineering

Led sensor-based data acquisition on DC-DC converters for predictive maintenance. Engineered meta-statistical features over sliding time windows, applied classical ML models (scikit-learn, GridSearch-tuned) for anomaly detection, and deployed compressed CNNs on resource-constrained TinyML devices. Conducted Bachelor's thesis at the lab.

October 2021 – February 2022

Dresden, Saxony

Internship – IoT Edge Computing & Machine Learning (AWS) · T-Systems Multimedia Solutions GmbH

Independently designed and deployed scalable cloud services and data engineering pipelines using AWS CDK. Contributed to predictive maintenance projects and gained hands-on SCRUM and CI/CD experience.

October 2020 – July 2021

Deggendorf, Bavaria

Research Assistant – Embedded Systems · DIT Project Lab for Hardware-Related Digitization

Developed a GUI application for sensor communication via UART, SPI, and I²C protocols using the PICkit Serial Analyzer. Bridged hardware and software integration in embedded systems projects.

Academia

Education

April 2023 – March 2026

Erlangen, Bavaria

M.Sc. Artificial Intelligence · Friedrich-Alexander-University (FAU)

Grade 1.3 — Focus on Advanced Deep Learning, Pattern Recognition, Statistical Learning, and Diffusion & Attention Models. Master thesis: "K-space Latent Diffusion for Accelerated MRI" (IdeaLab) — Grade 1.3.

October 2019 – March 2023

Deggendorf, Bavaria

B.Eng. Applied Computer Science · Deggendorf Institute of Technology (DIT)

Grade 1.4 — Focus on Embedded Systems. Bachelor thesis: "Generation, Optimization, and Evaluation of ML/DL Models for Tiny Machine Learning (TinyML)" — Grade 1.3.

August 2022

Odense, Denmark

Summer School – Deep Learning · University of Southern Denmark (SDU)

Completed a hands-on project classifying healthy vs. pneumonia-affected lungs using convolutional neural networks.

Selected work

Projects

Research and engineering work I'm proud of.

K-space Latent Diffusion for Accelerated MRI

Master thesis (IdeaLab, FAU) — Grade 1.3. Proposed the kLD-MRI framework, applying latent diffusion models directly in k-space for accelerated MRI reconstruction. Designed k-space autoencoders preserving high-frequency structure in the latent domain, plus a consistency-guidance sampling method for undersampled reconstruction. Fully open-sourced for reproducible follow-on research.

PyTorchDiffusion ModelsMRI ReconstructionDeep LearningPython

SMMetrDE – Segmentation-guided Monocular Metric Depth Estimation

Framework combining semantic segmentation with monocular depth estimation for object-wise metric prediction. Integrated and benchmarked MonoDepth2, MiDaS, DPT, and ZoeDepth (RMSE, AbsRel, inference time) on KITTI dataset subsets, working hands-on with transformer-based vision models (ViT, BEiT, Swin).

PyTorchComputer VisionViTDepth EstimationPython

AI Agent for Kalah – Tournament Challenge

Competitive game-playing agent using minimax search, alpha-beta pruning, and heuristic evaluation to optimize gameplay decisions under strict time constraints.

PythonSearch AlgorithmsAIGame Theory

Deep Learning from Scratch

Core deep learning architectures implemented from scratch in Python — forward/backward passes for FNNs, CNNs, and RNNs, alongside custom optimizers (SGD, Adam), activation functions, and loss functions.

PythonPyTorchDeep LearningNeural Networks

Writing

From the blog

Notes on AI agents, machine learning, and building things end to end.

Jul 29, 2026

Why I Moved My Portfolio to Cloudflare Pages

Static hosting with edge functions, zero cold starts, and a free tier that actually covers a portfolio — here's what the migration looked like.

Jul 15, 2026

Lessons from Taking LLM Agents to Production

What changes when an agent leaves the notebook: orchestration graphs over loops, observability from day one, and why most 'agent bugs' are actually tool-design bugs.

All posts

Contact

Get in touch

Open to new opportunities, collaborations, or just a chat.