08.06.2026 aktualisiert
verifiziert
Premiumkunde
100 % verfügbarAI Engineer | High-Performance & Scalable Software Engineering | Ex-Google SOC Mentor
Bischberg, Deutschland
Weltweit
Computer Science / M. Sc. - Ph.D. (expected 2026)Über mich
Software-Architekt & Performance-Spezialist mit 15+ Jahren Erfahrung. Skalierung, Cloud-native Architekturen, Echtzeit-Systeme & Legacy-Modernisierung. Java, Python, C/C++, Elixir | AWS/Azure/GCP, Kubernetes | Certified Scrum Master. Ich liefere – nicht nur Folien.
Skills
AI AgentsMLOpsKünstliche IntelligenzAlgorithmus-EntwurfBoost (C++ Libraries)Cloud-EngineeringComputational EngineeringCUDAPythonMachine LearningTensorFlowAzure Machine LearningHigh Performance ComputingFeature-EngineeringData Science
AI/ML Research meets Production Engineering
Modern AI creates hard new problems: models that memorise private training data, inference systems that can't run at the edge, and LLMs where "just unlearn it" costs 50% of capability. That's where I work.
I'm a published researcher (ICPR 2026, NeurIPS 2026 submissions) and a senior engineer with 15+ years of production experience. I close the gap between research and shipping—whether that's building an ML pipeline, hardening a model against membership inference, or rethinking how inference runs on constrained hardware.
Research background (FAU Erlangen-Nürnberg)
• Machine Unlearning for LLMs: developed LISU, a composable layer-selective unlearning method achieving 96–98% forgetting effectiveness at 2–4% better utility than full gradient ascent. Works with any unlearning objective (GA, NPO, GD). ICPR 2026.
• Privacy Auditing: kNN-ActDiv, a per-sample membership inference attack using activation divergence across transformer layers. No shadow models. Outperforms Min-K% AUC on non-monotonic architectures. NeurIPS 2026 submission.
• Scaling LLM influence: systematic study of gradient-based influence (KFAC) across 8 LLMs (82M–8B). Four-regime taxonomy that predicts which unlearning strategy wins from the layer-influence shape alone. NeurIPS 2026 submission.
• Edge AI Runtime: Rust-based data-aware inference runtime applying HPC scheduling principles to edge ML. 3.3× faster than PyTorch, 62× faster than Ray, <50MB footprint. IPDS 2026.
How I can help
• Design and implement ML pipelines, AI backends, and LLM-integrated systems
• Privacy and security auditing for ML models (MIA, unlearning, compliance)
• Edge and distributed ML inference—latency-critical, resource-constrained environments
• Modernise and scale architectures (cloud-native, microservices, event-driven)
• Optimise performance-critical code (HPC background: C++, CUDA, distributed systems)
• Bridge research and product: from paper to production-grade implementation
Technical expertise
AI/ML: PyTorch, Hugging Face Transformers, QLoRA/LoRA, Influence Functions, Activation Analysis, Scikit-learn, TensorFlow, LangChain, OpenAI API, MLOps, Databricks
Systems/HPC: Rust, C/C++ (C++17/20), CUDA/OpenMP/MPI, distributed computing, edge inference
Backend: Python/FastAPI, Java/Spring Boot, Elixir/OTP, Node.js/TypeScript, PostgreSQL, MongoDB, Kafka, Neo4j
Cloud/Infra: AWS, Azure, GCP, Kubernetes, Docker, Terraform, Ansible, GitLab CI/CD
Frontend: React, Vue.js, Angular, TypeScript
Let's connect
If you're building AI systems, need ML privacy expertise, or want research-grade engineering applied to real products—let's talk.
Sprachen
DeutschMutterspracheEnglischMuttersprache
Projekthistorie
Google Summer of Code (GSoC) Mentor – Google
Duration: 2016 - 2017 | Google LLC, Mountain View, CA
As a GSoC Mentor, I guided contributors in open-source development, ensuring high-quality, scalable solutions. My role included defining project scopes, reviewing proposals, conducting code reviews, and providing technical mentorship in collaboration with Google’s developer ecosystem. Through this, I helped shape the next generation of engineers while driving innovation in open-source projects.
Duration: 2016 - 2017 | Google LLC, Mountain View, CA
As a GSoC Mentor, I guided contributors in open-source development, ensuring high-quality, scalable solutions. My role included defining project scopes, reviewing proposals, conducting code reviews, and providing technical mentorship in collaboration with Google’s developer ecosystem. Through this, I helped shape the next generation of engineers while driving innovation in open-source projects.