Munich, Bavaria, Germany
I’m a full-stack AI engineer with 5+ years of experience building production-grade machine learning systems end-to-end. My work spans LLM engineering, agentic workflows, multimodal ML, computer vision, embeddings, retrieval, and efficient on-device inference. I thrive when taking ideas from data → model → API → deployment and turning them into real, usable AI products. At BMW, I design and deploy AI-driven systems that accelerate design workflows, automate repetitive processes, and enable teams to explore concepts faster using computer vision, representation learning, and intelligent retrieval. My role sits at the intersection of engineering and product — understanding creative workflows, identifying bottlenecks, and building practical AI tools that meaningfully improve speed and quality. In parallel, I’ve built several independent, ML-first products from scratch, including: 1. Agentic AI assistants with multi-step reasoning and tool-use 2. RAG-based research systems for fast knowledge extraction 3. OCR + detection pipelines using YOLO and transformer-based models 4. Mobile learning apps powered by multimodal AI and step-by-step reasoning 5. Food-distribution platforms using predictive modeling and workflow automation In each project, I own the full lifecycle: data engineering, model development and fine-tuning, evaluation, optimization (quantization/distillation), API design, backend infrastructure, deployment, and iterative improvements based on real user feedback. 🧰 Core Capabilities 1. ML & LLM Engineering: Transformers, agentic systems, RAG, PEFT, embeddings, CV 2. Full-Stack AI: FastAPI, Python, vector databases, backend systems, orchestrated workflows 3. MLOps & Infra: Docker, MLflow, scalable serving, monitoring, cloud deployment 4. Efficient / On-Device AI: quantization, pruning, distillation, CPU/GPU-optimized inference 5. Product & Strategy: rapid prototyping, UX-focused design, AI product lifecycle ownership 6. Cross-Functional Leadership: translating ambiguous ideas into clear technical roadmaps I care deeply about building AI systems that work in the real world — fast, reliable, explainable, and aligned with how people actually solve problems. Whether it’s designing an agentic workflow, optimizing a model for edge devices, or turning a vague idea into a deployed AI product, I love the entire process. If you’re building something ambitious in applied AI, I’d love to connect.
Taking AI systems to production
This is an initiative that involves volunteering, pro-bono work as well as hobby projects! - AI for high-frequency trading — improved returns and reduced volatility by combining LLMs, sentiment analysis, and proprietary signals, with strong compliance and model validation. - AI for autonomous agents — developed multi-agent LLM systems capable of planning, reasoning, and task execution across tools and APIs, enabling automation of complex workflows. - AI for personalised recommendations — built a real-time recommendation engine using deep learning and user embeddings, improving engagement and conversion for digital platforms. - AI for fraud detection — applied graph neural networks and anomaly detection to identify transactional fraud patterns in financial data, increasing detection accuracy while reducing false positives. - AI for multimodal understanding — engineered a system combining vision and language models (e.g., CLIP, BLIP) to extract structured insights from product images and descriptions, powering smarter search and classification. - AI for synthetic data generation — designed a generative pipeline to produce domain-specific synthetic datasets using diffusion models and LLMs, reducing data acquisition bottlenecks and improving model robustness. - AI for customer feedback analysis — implemented NLP pipelines to extract insights from large-scale support tickets and reviews, enabling data-driven product decisions. - AI for predictive maintenance — applied time-series forecasting and anomaly detection on sensor data to reduce downtime and optimise maintenance schedules for industrial equipment. - AI for document intelligence — developed LLM-based systems to extract structured insights from contracts, invoices, and reports, accelerating workflows in legal and financial services.
AI for the Automotive Industry - Working on a lot of interesting AI use cases, building AI and agentic systems
AI for the insurance industry - building RAG-based assistants to automate complex compliance processes from identification, verification, to reporting from weeks to minutes.