About

About John Pazarzis

I am a system architect, expert programmer and researcher focused on high-performance computing, GPU-accelerated algorithms, and architecture for performance-critical systems.

Professional Evolution

My career spans enterprise financial systems and large-scale distributed infrastructure. I have worked across organizations ranging from startups to large enterprises, including Google.

Over the years, I have built deep expertise in Linux systems and in languages including C, C++, C#, Java, and Python. Whether solving high-frequency trading problems in C++ or designing distributed systems and networked architectures, I approach each project with an emphasis on low-level control, reliability, and architectural clarity.

Engineering Philosophy

I believe software engineering begins with architecture, not tools. My process starts with structural blueprints and architectural documentation long before implementation begins. This discipline helps prevent the kind of “spaghetti code” that can result from over-reliance on AI-generated code.

Architecting for Silicon

My approach to programming has undergone a rigorous evolution. Where I once viewed software primarily through the lens of high-level abstractions—focusing on logic and developer velocity—I have transitioned to a “hardware-aware” model of engineering.

Mastering CUDA and GPU-accelerated computing did more than add a new tool to my stack; it fundamentally altered how I visualize the execution of code. I no longer treat the machine as a “black box” that executes my logic. Instead, I design with a “white box” mindset: visualizing memory hierarchies, bank conflicts, and register pressure before a single line of code is written. This shift has been transformative, refining my work by allowing me to align software structures with the underlying hardware, effectively moving me from a builder of application logic to an architect of computational performance.

Technical Expertise

My current primary languages are C, C++, CUDA, and Python. I bring deep expertise in Machine Learning, including TensorFlow, Convolutional Neural Networks (CNNs), Random Forests, and Support Vector Machines (SVMs). These technologies form the core of my high-performance and research work, while C# and Java remain part of my broader professional background.

Current Focus

Today, my work is driven by the performance limits of hardware and the engineering of sophisticated AI systems. I am currently focused on three areas:

  • Biotech & Medical AI: Developing GPU-accelerated algorithms for biological sequence data through LocalX, and building machine-learning pipelines for MRI-based dementia prediction.
  • Custom LLM Systems: As an early adopter of LLM technologies, I design and build robust, high-performance systems, including end-to-end custom RAG (Retrieval-Augmented Generation) pipelines for specialized data domains.
  • Architecture-First Engineering: Using AI as a tool to support design and implementation, while keeping system architecture explicit, maintainable, and human-directed.

Research & Intellectual Property

My interest in systems design also extends to formal research and patentable innovation. My work explores bridge-pattern architectures for hybrid Python/C/CUDA environments and privacy-preserving statistical aggregation.

About This Blog

I started this blog in 2013. It serves as a living record of my evolving technical philosophy and a place to track how my views on software development, architectural patterns, and the industry in general have changed over time.