I'm a cross-disciplinary AI engineer with a background in cognitive science, signal processing, and applied statistics. My professional work has spanned industrial computer vision (wood defect detection, conveyor-belt inspection on embedded cameras), energy market analysis and financial risk modeling, and neuroscience research at the Max Planck Institute. What these contexts share is messy real-world data, constrained hardware, and stakeholders who need plain-language explanations of what the AI is actually doing — and that intersection is where I do my best work. My strongest asset is moving quickly from an ill-defined problem to a working prototype. At one employer, I identified their hardest unsolved vision problem at a job fair, built an object detection prototype in six months, and was hired to scale it into production. I've done similar things with portfolio risk models, LLM-based document pipelines, and custom hardware projects. I work across the full pipeline: data acquisition and labeling strategy, model training and evaluation, quantization-aware optimization for edge deployment, firmware integration in Python, Lua, and C++, and user-facing documentation. More recently I've been building LLM-powered systems — RAG pipelines, embedding-based filtering, transcript summarization with context nesting — using both cloud APIs and local inference. What I bring beyond the technical stack is a strong intuition for where models break and why. My statistics training means I examine residuals, question distributional assumptions, and catch failure modes before they reach production. I also genuinely enjoy the human side: designing interfaces, writing clear documentation, and translating model behavior into decisions that non-technical people can act on. My electrical engineering background and years of hardware tinkering mean I'm comfortable reading schematics, talking to firmware teams, and understanding the physical constraints a model will eventually run inside. I'm best suited for projects that need rapid prototyping, cross-domain problem-solving, or a fresh perspective on a stuck technical problem.…I'm a cross-disciplinary AI engineer with a background in cognitive science, signal processing, and applied statistics. My professional work has spanned industrial computer vision (wood defect detection, conveyor-belt inspection on embedded cameras), energy market analysis and financial risk modeling, and neuroscience research at the Max Planck Institute. What these contexts share is messy real-world data, constrained hardware, and stakeholders who need plain-language explanations of what the AI is actually doing — and that intersection is where I do my best work. My strongest asset is moving quickly from an ill-defined problem to a working prototype. At one employer, I identified their hardest unsolved vision problem at a job fair, built an object detection prototype in six months, and was hired to scale it into production. I've done similar things with portfolio risk models, LLM-based document pipelines, and custom hardware projects. I work across the full pipeline: data acquisition and labeling strategy, model training and evaluation, quantization-aware optimization for edge deployment, firmware integration in Python, Lua, and C++, and user-facing documentation. More recently I've been building LLM-powered systems — RAG pipelines, embedding-based filtering, transcript summarization with context nesting — using both cloud APIs and local inference. What I bring beyond the technical stack is a strong intuition for where models break and why. My statistics training means I examine residuals, question distributional assumptions, and catch failure modes before they reach production. I also genuinely enjoy the human side: designing interfaces, writing clear documentation, and translating model behavior into decisions that non-technical people can act on. My electrical engineering background and years of hardware tinkering mean I'm comfortable reading schematics, talking to firmware teams, and understanding the physical constraints a model will eventually run inside. I'm best suited for projects that need rapid prototyping, cross-domain problem-solving, or a fresh perspective on a stuck technical problem.WWWW…