Linux & Virtualization Engineering Portfolio
Linux and virtualization engineer documenting real delivery patterns: clear issue statements, implementation choices, and production outcomes.
A production-grade personal finance application serving real users. Features an AI-powered financial coach, offline-first architecture, and cross-platform syncing.
RHEL lifecycle, automation, virtualization, and production operations.
Ansible playbooks, task automation, and configuration management.
Applied AI across cloud services, local inference, and practical delivery lessons.
Azure architecture, infrastructure design, and delivery practices.
Running models on local hardware with privacy-first workflows.
Kotlin projects, notes, and engineering experiments.
Quick commands and reusable building blocks for day-to-day work.
How splitting a monolithic system prompt into static and per-session layers improved estimated KV cache hit rates from ~42% to ~76% and reduced input costs by an estimated 57% on a Firebase Functions app running DeepSeek V4 Flash.
Full benchmark data on running MoE and dense LLMs on AMD consumer hardware — quantization comparison, power cap analysis, KV cache tuning, and context limits on 16GB VRAM.
A practical guide to splitting an oversized Git PR into clean, topic-focused branches using path-based checkout from a fresh branch off main.
Benchmarked every viable local LLM (350M to 26B, CPU and NPU) through a live Discord agent pipeline on RK3588. Found NPU beats CPU at same quality, code is solved at any size, and 4B+ models are slower AND worse than 2B on this board.
Publishing a practical local-AI control plane for llama.cpp: remote model loading, runtime tuning, streaming chat, and real REAP model serving on a Radxa ROCK 5B+.
An advanced benchmark report for running Qwen3.5 locally on RK3588 with source-built llama.cpp: prefill speed, decode speed, stable context, tool-calling behavior, and the practical model choices that actually work on a Radxa ROCK 5B+.