Bridging Research & Production

I am a Machine Learning Research Scientist focused on efficient AI — turning large vision and language models into fast, robust, production-ready systems. My work spans model optimization (quantization, pruning, sparsification), accelerator-aware inference on NPUs and GPUs, and real-time pipelines for edge and cloud.

I currently work on generative AI and computer vision at a global optics and optoelectronics company in Munich, and conduct independent research on efficient AI in affiliation with leading European research institutes, collaborating with academic and industry partners on EU-funded projects.

Quick Facts

• 📍 Based in Munich, Germany
• 🎓 MSc in Robotics, Cognition, Intelligence — Technical University of Munich, with a research stay at EPFL
• 🌟 Published at NeurIPS 2024, EurIPS 2025, and AABI on neural network compression
• 🧪 Reviewer for ICML, NeurIPS, and JAIR
• 🔧 Foundation models, computer vision, LLM optimization, Bayesian inference, edge AI
• 🌍 Arabic & French (native), English & German (fluent)

What Drives Me

The best model is not the biggest one — it's the one that solves the problem within real constraints. I thrive on the gap between state-of-the-art research and deployment: compressing vision foundation models for real-time edge inference, running LLMs on-device, and developing principled Bayesian approaches to network sparsification. I care about research that ships and delivers measurable value.

Experience Highlights

• Zeiss — Machine Learning Research Scientist: generative AI & computer vision on edge accelerators
• INSAIT / ETH Zurich ecosystem — Independent researcher on efficient AI, with a first-author submission at NeurIPS 2026
• CARIAD (Volkswagen Group) — Instance & semantic segmentation research for vehicle perception
• ETH Zurich — NeurIPS 2024 publication on Bayesian sparsification
• BMW — Neural network optimization for resource-constrained automotive hardware
• Huawei — RL & LLM-based optimization for robotic fleets
• Intel — Sparse & quantized networks (4× memory reduction, up to 7× speedup)
• EPFL — Saliency prediction research
• Infineon — AI pipelines for microchip testing & anomaly detection

Certifications