Bernhard Specht
Research Engineer
I turn hard problems into shipped products. From brain-computer interfaces to privacy-preserving ML systems and beyond — I research, build, and deliver end-to-end.
I turn hard problems into shipped products. From brain-computer interfaces to privacy-preserving ML systems and beyond — I research, build, and deliver end-to-end.
MyelinZ / Myelin-H
A hybrid brain-computer interface platform for remote neurological monitoring and rehabilitation. Built the full product from scratch — backend, data pipeline, real-time sensor-driven game engine, clinical dashboard, and mobile app. Turning neurological paradigms and clinical tests into interactive games, adopted and tested by major clinics.
A selection of demos from the BodyMirror platform.
Side Project
Snap a photo of a dish. Describe a meal in your own words. CrunchGuide turns your ideas into complete recipes — with ingredients, steps, and full nutrition data.
Recipe Creation
My Recipes
Recipe Details
Meal Planning
Take a photo of something you ate, dictate a recipe from memory, or type a quick idea. CrunchGuide builds a structured recipe with ingredients, portions, and step-by-step instructions.
Every recipe comes with detailed nutrition info — powered by USDA food data. Track calories, protein, carbs, fat, fiber, vitamins, minerals, amino acids, and more.
Drag and drop recipes into your weekly schedule. Stay on top of your eating goals without the guesswork.
Generate shopping lists from your meal plans automatically. Check items off as you shop — nothing forgotten, nothing wasted.
Share recipes, meal plans, and shopping lists with family or friends. Plan dinner together or split the grocery run.
Save, organize, and revisit your favorite recipes anytime. Build a personal cookbook that grows with you.
Technical University of Munich
Brain-computer interface projects from my time as a teaching assistant and working student.
Ph.D. Research
A privacy-preserving federated learning system for cross-device medical data. Raw medical data never leaves patients' mobile devices — the system uses secret sharing to enable collaborative model training without exposing individual records.
