{"@attributes":{"version":"2.0"},"channel":{"title":"\u00d6zkan Pakdil Software Engineer","link":"https:\/\/ozkanpakdil.github.io\/","description":"Recent content on \u00d6zkan Pakdil Software Engineer","image":{"title":"\u00d6zkan Pakdil Software Engineer","url":"https:\/\/ozkanpakdil.github.io\/images\/papermod-cover.png","link":"https:\/\/ozkanpakdil.github.io\/images\/papermod-cover.png"},"generator":"Hugo -- 0.148.2","language":"en","copyright":"\u00d6zkan Pakdil","lastBuildDate":"Sun, 19 Jul 2026 00:00:00 +0000","item":[{"title":"How to add NVIDIA free models to VS Code","link":"https:\/\/ozkanpakdil.github.io\/posts\/my_collections\/2026\/2026-07-19-how-to-add-nvidia-free-model\/","pubDate":"Sun, 19 Jul 2026 00:00:00 +0000","guid":"https:\/\/ozkanpakdil.github.io\/posts\/my_collections\/2026\/2026-07-19-how-to-add-nvidia-free-model\/","description":"

NVIDIA offers free access to their powerful Nemotron models through their NVIDIA Build platform<\/a>. In this post, I’ll walk you through how to get a free API key and configure it in VS Code to use models like nvidia\/nemotron-3-ultra-550b-a55b<\/code> directly in VS Code. Also you can choose any other model from here<\/a>.<\/p>\n

Prerequisites<\/h3>\n

VScode needs to be above 1.27 it needs to have customendpoint supporting<\/p>\n

Getting Your Free NVIDIA API Key<\/h2>\n
    \n
  1. Go to build.nvidia.com<\/a><\/li>\n
  2. Sign in or create an NVIDIA account<\/li>\n
  3. Navigate to the model you want to use (e.g., Nemotron 3 Ultra<\/a>)<\/li>\n
  4. Click “Get API Key” button<\/li>\n<\/ol>\n

    Note<\/strong>: In my experience, the “Get API Key” button in the UI didn’t work properly (it appeared to do nothing). If this happens to you, here’s how to get your API key:<\/p>"},{"title":"Beyond Functionality: Why Code Hygiene is Your Project's Immune System","link":"https:\/\/ozkanpakdil.github.io\/posts\/my_collections\/2026\/2026-04-22-code-hygiene\/","pubDate":"Wed, 22 Apr 2026 00:00:00 +0000","guid":"https:\/\/ozkanpakdil.github.io\/posts\/my_collections\/2026\/2026-04-22-code-hygiene\/","description":"

    In the world of software engineering, we often obsess over the “Testing Pyramid.” We pour resources into unit tests, integration tests, and E2E suites. These are critical, they tell us that our features work as designed. But there\u2019s a shadowy category of bugs that traditional tests often miss: the architectural “anti-patterns” and “API misuses” that don’t break functionality today but lead to system failures, memory leaks, or portability issues tomorrow.<\/p>"},{"title":"Accelerating LLMs on Debian 13: Setting up Vulkan for llama.cpp","link":"https:\/\/ozkanpakdil.github.io\/posts\/my_collections\/2026\/2026-03-22-vulkan-llamacpp-debian-13\/","pubDate":"Sun, 22 Mar 2026 09:47:00 +0300","guid":"https:\/\/ozkanpakdil.github.io\/posts\/my_collections\/2026\/2026-03-22-vulkan-llamacpp-debian-13\/","description":"

    After setting up CUDA on my other laptop, I moved to a different(older) machine that doesn’t have an NVIDIA GPU. This one is an everyday laptop with integrated Intel graphics, but that doesn’t mean we have to settle for slow CPU-only performance.<\/p>\n

    On this machine, I switched to the Vulkan<\/strong> backend for llama.cpp<\/code> and the results were even more dramatic than I expected.<\/p>\n

    Machine Hardware Info<\/h3>\n

    This laptop is running Debian 13 (Trixie\/Sid)<\/strong> with the following specs:<\/p>"},{"title":"Accelerating LLMs on Debian 13: Setting up CUDA for llama.cpp","link":"https:\/\/ozkanpakdil.github.io\/posts\/my_collections\/2026\/2026-03-20-cuda-llamacpp-debian-13\/","pubDate":"Fri, 20 Mar 2026 01:35:00 +0300","guid":"https:\/\/ozkanpakdil.github.io\/posts\/my_collections\/2026\/2026-03-20-cuda-llamacpp-debian-13\/","description":"

    Setting up NVIDIA CUDA on Debian 13 (Trixie\/Sid) to run Large Language Models (LLMs) can be a bit of a journey, especially if you’re transitioning from the default open-source drivers to the proprietary stack required for GPGPU workloads.<\/p>\n

    Over the last few days, I’ve been working on getting llama.cpp<\/code> to run with CUDA on my laptop to see how much of a difference it makes compared to pure CPU execution.<\/p>"},{"title":"Tuning Podman on macOS to Match OrbStack Performance","link":"https:\/\/ozkanpakdil.github.io\/posts\/my_collections\/2026\/2026-03-08-tuning-podman-macos-performance\/","pubDate":"Sun, 08 Mar 2026 22:02:00 +0300","guid":"https:\/\/ozkanpakdil.github.io\/posts\/my_collections\/2026\/2026-03-08-tuning-podman-macos-performance\/","description":"

    \n

    Note:<\/strong> These are suggested optimizations I have not personally tried yet. I’m blogging them as I’m planning to test them throughout this week.<\/p><\/blockquote>\n

    OrbStack is highly optimized for macOS, using a proprietary, high-performance networking stack and a custom VirtioFS implementation with aggressive caching. Podman, while being open-source and standard-compliant, can be tuned to significantly bridge the performance gap.<\/p>\n

    The following plan outlines key areas where Podman’s performance can be improved on macOS:<\/p>"},{"title":"Atlassian MCP","link":"https:\/\/ozkanpakdil.github.io\/posts\/my_collections\/2026\/2026-02-08-atlassian-mcp\/","pubDate":"Sun, 08 Feb 2026 00:00:00 +0000","guid":"https:\/\/ozkanpakdil.github.io\/posts\/my_collections\/2026\/2026-02-08-atlassian-mcp\/","description":"

    I have been using Atlassian MCP<\/a> with internal Confluence and Jira, and it has been wonderful.<\/p>\n

    Finding internal information is often challenging and time-consuming. To be honest, searching through Jira or Confluence and locating the right information can be really difficult.<\/p>\n

      \n
    1. Create Jira and Confluence API tokens from your internal site profile page. For example: https:\/\/internalconfluence.company.com\/profile\/personal<\/code> for Confluence and https:\/\/jira.company.com\/secure\/admin\/CreateAPIToken!default.jspa<\/code> for Jira. These URLs may vary depending on your setup.<\/li>\n
    2. Create an mcp.json<\/code> file in the .vscode<\/code> folder for Visual Studio Code, or place this MCP configuration in the appropriate folder for your IDE of choice:<\/li>\n<\/ol>\n
      {\n<\/span><\/span>  "mcpServers"<\/span>: {\n<\/span><\/span>    "mcp-atlassian"<\/span>: {\n<\/span><\/span>      "command"<\/span>: "uvx"<\/span>,\n<\/span><\/span>      "args"<\/span>: ["mcp-atlassian"<\/span>],\n<\/span><\/span>      "env"<\/span>: {\n<\/span><\/span>        "JIRA_URL"<\/span>: "https:\/\/jira.company.com"<\/span>,\n<\/span><\/span>        "JIRA_USERNAME"<\/span>: "your.email@company.com"<\/span>,\n<\/span><\/span>        "JIRA_API_TOKEN"<\/span>: "your_api_token"<\/span>,\n<\/span><\/span>        "CONFLUENCE_URL"<\/span>: "https:\/\/internalconfluence.company.com\/wiki"<\/span>,\n<\/span><\/span>        "CONFLUENCE_USERNAME"<\/span>: "your.email@company.com"<\/span>,\n<\/span><\/span>        "CONFLUENCE_API_TOKEN"<\/span>: "your_api_token"<\/span>\n<\/span><\/span>      }\n<\/span><\/span>    }\n<\/span><\/span>  }\n<\/span><\/span>}\n<\/span><\/span><\/code><\/pre><\/div>
      Remember to run podman-desktop or docker desktop. Because this MCP works as a docker container.<\/p>"},{"title":"Building a Real-time File I\/O Heatmap with eBPF and Java 25","link":"https:\/\/ozkanpakdil.github.io\/posts\/my_collections\/2026\/2026-01-21-ebpf-java-heatmap\/","pubDate":"Wed, 21 Jan 2026 08:52:00 +0300","guid":"https:\/\/ozkanpakdil.github.io\/posts\/my_collections\/2026\/2026-01-21-ebpf-java-heatmap\/","description":"
      Have you ever wondered exactly which files are being hammered by your Linux system in real-time? While tools like iotop<\/code> or lsof<\/code> are great, sometimes you want something more visual, custom, and lightweight.<\/p>\n
      In this post, I’ll walk you through how I built a Real-time File I\/O Heatmap<\/strong> using the power of eBPF<\/strong> for data collection and Java 25<\/strong> for a modern Terminal UI (TUI).<\/p>\n
      What is eBPF and Why Use It?<\/h3>\n
      eBPF (Extended Berkeley Packet Filter) is a revolutionary technology that allows you to run sandboxed programs in the Linux kernel without changing kernel source code or loading kernel modules.<\/p>"},{"title":"Bun Joins the Microservice Framework Benchmark: Surprisingly Fast JavaScript Runtime","link":"https:\/\/ozkanpakdil.github.io\/posts\/my_collections\/2026\/2026-01-10-bun-microservice-framework-benchmark\/","pubDate":"Sat, 10 Jan 2026 17:00:00 +0000","guid":"https:\/\/ozkanpakdil.github.io\/posts\/my_collections\/2026\/2026-01-10-bun-microservice-framework-benchmark\/","description":"
      Introduction<\/h2>\n
      Today I’m excited to announce the addition of Bun<\/strong> to our microservice framework benchmark suite<\/a>. The results are nothing short of remarkable . Bun has proven to be one of the fastest runtimes in our entire test suite, competing directly with Rust frameworks!<\/p>\n
      What is Bun?<\/h2>\n
      Bun is a modern JavaScript runtime built from scratch using Zig<\/a> and JavaScriptCore<\/a> (the engine that powers Safari). It’s designed to be a drop-in replacement for Node.js with a focus on:<\/p>"},{"title":"Eclipse Collections vs JDK Collections: A Performance Deep Dive","link":"https:\/\/ozkanpakdil.github.io\/posts\/my_collections\/2025\/eclipse-collections-vs-jdk-collections\/","pubDate":"Mon, 29 Dec 2025 00:00:00 +0000","guid":"https:\/\/ozkanpakdil.github.io\/posts\/my_collections\/2025\/eclipse-collections-vs-jdk-collections\/","description":"Comparing the performance of Eclipse Collections against standard JDK Collections using JMH benchmarks. Spoiler: EC is slightly faster!"},{"title":"From PKIX errors to a clean mTLS + Feign + IAM demo","link":"https:\/\/ozkanpakdil.github.io\/posts\/my_collections\/2025\/mtls-feignclient-client-server-iam\/","pubDate":"Fri, 05 Dec 2025 21:50:00 +0300","guid":"https:\/\/ozkanpakdil.github.io\/posts\/my_collections\/2025\/mtls-feignclient-client-server-iam\/","description":"A minimal, runnable client\u2013server example that fixes PKIX issues in OpenFeign with mutual TLS and adds application\u2011level identity\/authorization via Spring Security X.509."},{"title":"Adding Golang and Express.js to the Microservice Framework Benchmark Suite","link":"https:\/\/ozkanpakdil.github.io\/posts\/my_collections\/2025\/2025-12-03-microservice-framework-test-golang-expressjs\/","pubDate":"Wed, 03 Dec 2025 21:49:00 +0000","guid":"https:\/\/ozkanpakdil.github.io\/posts\/my_collections\/2025\/2025-12-03-microservice-framework-test-golang-expressjs\/","description":"
      Introduction<\/h2>\n
      Test results for this benchmark run \u2192<\/a><\/p>\n
      Over the last two days, I’ve expanded our microservice framework benchmark suite to include two new contenders: Golang<\/strong> and Express.js<\/strong>. This addition allows us to compare performance across a broader spectrum of technologies, from compiled languages like Rust and Go to JVM-based frameworks and Node.js.<\/p>\n
      New Additions<\/h2>\n
      Golang (Go 1.24.10)<\/h3>\n
      Go was added using the standard library only<\/strong> - no external frameworks. The implementation uses net\/http<\/code> package which is known for its excellent performance and simplicity.<\/p>"},{"title":"How to solve macos hdmi sound control problem","link":"https:\/\/ozkanpakdil.github.io\/posts\/my_collections\/2025\/21-11-macos-hdmi-soundissue-fix\/","pubDate":"Fri, 21 Nov 2025 00:00:00 +0000","guid":"https:\/\/ozkanpakdil.github.io\/posts\/my_collections\/2025\/21-11-macos-hdmi-soundissue-fix\/","description":"
      So I got my Mac Mini M4 from Amazon for \u00a3500 and started using it. I had so many problems with the shortcuts I normally use  even Ctrl+A wasn\u2019t working, I had to use Win+A, and many other shortcuts were different. One of the biggest problems was using the sound keys on the keyboard. On Linux they worked fine: sound up and down controlled the output volume. But on macOS it didn\u2019t work. Very strange policy Apple has macOS doesn\u2019t allow the user to control end devices connected through HDMI.<\/p>"},{"title":"MacOS on debian QEMU KVM","link":"https:\/\/ozkanpakdil.github.io\/posts\/my_collections\/2025\/13-11-macos-kvm\/","pubDate":"Thu, 13 Nov 2025 00:00:00 +0000","guid":"https:\/\/ozkanpakdil.github.io\/posts\/my_collections\/2025\/13-11-macos-kvm\/","description":"
      From Frustration to Breakthrough: Running macOS on KVM<\/h1>\n
      For years, I chased the dream of running macOS in a virtual machine<\/strong>.
      \nOn Windows, I tried VMware and VirtualBox countless times with different tutorials and blogs. Each attempt ended in frustration: crashes, unsupported hardware, endless configuration rabbit holes. It felt like a goal always just out of reach. And finally I found       https:\/\/github.com\/kholia\/OSX-KVM<\/a> it has the readme which explains the steps for setting up.<\/p>"},{"title":"Paint, a lightweight image editor for quick edits","link":"https:\/\/ozkanpakdil.github.io\/posts\/my_collections\/2025\/paint-lightweight-image-editor\/","pubDate":"Wed, 12 Nov 2025 00:00:00 +0000","guid":"https:\/\/ozkanpakdil.github.io\/posts\/my_collections\/2025\/paint-lightweight-image-editor\/","description":"
      Why I built Paint<\/h2>\n
      Whenever I wanted to make a very small edit on my Debian laptop, crop a screenshot, add a quick arrow, or block out a small area, the system only had GIMP for editing images. Powerful, but heavy and slow for tiny, frequent tasks. I wanted something nimble: quick to open, easy to use, and focused on the common one-off edits people do dozens of times a day.<\/p>"},{"title":"Java imperative vs functional in 2025 - revisiting a 2015 microbenchmark","link":"https:\/\/ozkanpakdil.github.io\/posts\/my_collections\/2025\/java-imperative-vs-functional-2025\/","pubDate":"Mon, 29 Sep 2025 20:22:00 +0000","guid":"https:\/\/ozkanpakdil.github.io\/posts\/my_collections\/2025\/java-imperative-vs-functional-2025\/","description":"
      Quick numbers (avg; smaller is faster)<\/p>\n