I’m a self-directed technical problem solver with an enterprise IT service-desk background. I use Linux, software, experimentation, research, and deliberately structured AI-assisted workflows to investigate problems that cross application, infrastructure, GPU, video, and agentic-system boundaries.
My strongest work is not only the software I built. It is the progression in how I solve difficult problems: clear goals, explicit evidence, model specialization, independent evaluation, adversarial review, falsifiable experiments, remediation loops, and human responsibility for judgment and verification.
Flagship experimental R&D adapting AMD FSR 4.1-style temporal reconstruction/upscaling to ordinary decoded video while preserving the original frame cadence. The project investigates which renderer-derived temporal signals can be recovered from finished video, which must be synthesized, which cannot be trusted, and which inputs actually change reconstruction behavior.
AMD-first Linux AI video-enhancement application built around C++, ROCm, MiGraphX, HIP, Vulkan, and FFmpeg. Integration and performance limits in the underlying inference stack led to measured MiGraphX GPU optimizations, benchmarking, four current upstream ROCm pull requests, and integration of the resulting work back into the application.
Evidence-driven reverse engineering of AMD FSR 4.1’s neural temporal upscaler: shader and weight extraction, dispatch and resource-structure analysis, reproducible tooling, strict claim verification, and explicit boundaries between measured facts, inference, and behavior not directly observed.
- Human-defined goals, requirements, scope, and final judgment
- Digital-intelligence specialization for orchestration, implementation, evaluation, and research
- Independent review and adversarial challenge rather than trusting one model's first answer
- Measured evidence, provenance, uncertainty labels, and reproducible conditions
- Remediation loops, regression checks, and escalation only when the problem justifies it
- Preservation of negative results and failed experiments when they explain system behavior
AI assistance expands the range of problems I can attack. It does not replace responsibility for understanding, directing, testing, or defending the result.
Enterprise IT/service-desk troubleshooting, Linux and hardware, cloud infrastructure, C++, JavaScript/Node, Python, GPU/video systems, technical research, and self-directed software engineering.
Certifications: CompTIA A+, Microsoft Azure Fundamentals (AZ-900), Microsoft Azure AI Fundamentals (AI-900), and Linux Essentials.
Open to entry-level AI automation, agentic systems, technical R&D, systems integration, and advanced technical-support opportunities where rigorous AI-assisted problem solving is an advantage.