I architect distributed AI systems with 6+ years engineering multi-agent platforms, hybrid semantic search, and MLOps across healthcare, automotive, and logistics.
I'm a Senior AI Architect with 6+ years engineering distributed AI systems across healthcare, automotive, and logistics. I specialize in multi-agent systems, hybrid semantic search (BM25 + HNSW + RRF + HyDE), LLMOps, and scalable AI architectures on AWS, Azure, and GCP.
Currently I'm building AI-powered medical assistants for rare chronic diseases at mama health — a patient-side AI companion guiding treatment journeys and a pharma-side analytics assistant supporting partners including Bayer. Previously delivered enterprise AI to Porsche, BMW, DHL, Vodafone, Red Bull, and Siemens — generating €300K+ in annual savings through legacy modernization and multi-agent orchestration platforms.
Based in Berlin, I hold an M.Sc. in Computer Engineering from TU Berlin (Grade 1.4, Thesis 1.0). I'm multilingual: Arabic (native), English, German, Turkish, and French.
Master's Thesis (Grade 1.0): Combined LLMs with internet retrieval and uncertainty modeling to reduce hallucinations.
Industry 4.0 solution using eBPF & XDP for sub-millisecond detection latency across multi-cloud environments.
Multilingual chatbot for e-commerce, providing personalized book recommendations and order tracking.
Implementation of the BERT-based sequential recommendation system for personalized content suggestions.
Private cloud infrastructure using Kolla-Ansible on Google Cloud Platform with performance benchmarking.
Research on security vulnerabilities in AI-based operations, testing defense mechanisms against adversarial examples.
Exploring the path from LLMs to AGI through chain-of-thought reasoning, tree-of-thought exploration, and advanced techniques like MCTS and RLHF. A deep dive into building an artificial brain by simulating human cognitive processes with modular AI architectures.
A comprehensive analysis of Europe's path to digital independence. Covering EU-compliant cloud infrastructure, open-source alternatives, sovereign AI models, and the regulatory landscape shaping the future of European tech. From Mistral AI to Nextcloud, discover the ecosystem enabling true digital sovereignty.
I regularly write about AI/ML, system architecture, and emerging technologies. Follow me on Medium to stay updated.
Follow on MediumArchitecting a multi-agent medical assistant platform for rare chronic disease communities across EU & US, built on LiteLLM, Pydantic, Claude SDK, and DSPy. Leading a patient-side AI companion that builds disease-progression models from structured symptom data, plus a pharma-side analytics assistant supporting partners including Bayer. Designing Hybrid Search pipelines on Qdrant (BM25 + dense Gemini embeddings via HNSW, RRF fusion, HyDE) evaluated with F1, MRR, and nDCG.
Led a team building an Arbeitszeugnisgenerator on ServiceNow HRSD + Now Assist for a German Mittelstand enterprise. Initiated a multi-language AI voice assistant integrating Twilio with ServiceNow Now Assist Voice for a major Frankfurt company. Architected a multi-tenant GenAI solution targeting €50K/year operational savings across 5+ clients.
Delivered enterprise AI to Porsche, BMW, DHL, Vodafone, Red Bull, and Siemens. Project Lead on Porsche's vehicle diagnostics modernization (team of 10, 30K+ workshops in 16 languages, 150K+ daily requests, €300K/year savings, 99.996% behavioral parity). Co-architected a multi-agent orchestration platform on AWS later sold to BMW. Built a LangGraph agent harness orchestrating 50+ coding agents across 200+ DHL repos, cutting tech-debt resolution by 60%. Delivered a GDPR-compliant Medical Conversational Agent processing 10TB+ clinical data. Promoted to Senior in Sep 2024.
Led design and implementation of a semantic search solution with major research centers, deployed via CI/CD to private cloud supporting 1,000+ active users. Built an MLFlow pipeline training and serving 16+ AI models behind a secure JWT endpoint, cutting deployment time by 50%.
Built interactive web apps for the public IT sector visualizing global research datasets (React, D3.js, FastAPI). Formulated a custom Rasa + BERT pipeline for multilingual government queries achieving 15% higher intent accuracy vs Dialogflow.
Microsoft Azure
The Linux Foundation
Google Cloud
DeepLearning.ai
Microsoft
Amazon Web Services