<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:content="http://purl.org/rss/1.0/modules/content/"><channel><title>Deep Dives &amp; Weekend Projects</title><link>https://blog.lukasmay.com/</link><description>Recent content on Deep Dives &amp; Weekend Projects</description><generator>Hugo -- 0.161.1</generator><language>en-us</language><lastBuildDate>Fri, 06 Mar 2026 23:07:26 -0500</lastBuildDate><atom:link href="https://blog.lukasmay.com/index.xml" rel="self" type="application/rss+xml"/><item><title>Deep Learning</title><link>https://blog.lukasmay.com/deep-dives/deep-learning/</link><pubDate>Fri, 06 Mar 2026 23:07:26 -0500</pubDate><guid>https://blog.lukasmay.com/deep-dives/deep-learning/</guid><description>&lt;h2 id="intro"&gt;Intro&lt;/h2&gt;
&lt;p&gt;This is an attempt to cover what I know about DL to some degree. Some stuff is very skippable, and I don&amp;rsquo;t really remember everything that I put in here, so there might be some repeating, but not much.&lt;/p&gt;
&lt;h2 id="the-foundations-of-deep-learning"&gt;The Foundations of Deep Learning&lt;/h2&gt;
&lt;p&gt;To understand how machine learning models work, you have to completely discard the idea that it &amp;ldquo;understands&amp;rdquo; anything. A model does not read text or see images. At the absolute lowest level, a neural network is just a massively complex sequence of mathematical operations executed on silicon. To feed data into that silicon, we must first translate reality into a format that a GPU’s compute cores can process. That translation layer is the tensor.&lt;/p&gt;</description></item><item><title>AI Stack Part 1</title><link>https://blog.lukasmay.com/deep-dives/ai-stack-part-1/</link><pubDate>Mon, 16 Feb 2026 01:16:08 -0500</pubDate><guid>https://blog.lukasmay.com/deep-dives/ai-stack-part-1/</guid><description>&lt;h1 id="overview"&gt;Overview&lt;/h1&gt;
&lt;p&gt;I have been looking into self-hosting LLMs, and this is my attempt to put everything I&amp;rsquo;ve learned about the subject in one place (so I can stop forgetting things). Alongside that, I wanted to include information about the setup I use to self-host LLMs on my laptop and the steps I took to build and optimize it. While that will come in the future, as there are still some things I am changing, and this is long enough already, I removed some of those parts to put in the next section.&lt;/p&gt;</description></item><item><title>LazyCargo</title><link>https://blog.lukasmay.com/projects/lazycargo/</link><pubDate>Sun, 07 Dec 2025 22:25:58 -0500</pubDate><guid>https://blog.lukasmay.com/projects/lazycargo/</guid><description>&lt;h2 id="introduction"&gt;Introduction&lt;/h2&gt;
&lt;p&gt;This is meant to be an outline of what I found while reversing the &lt;strong&gt;LazyCargo&lt;/strong&gt; malware sample. This malware sample is one part of the five pipedream/INCONTROLLER malware framework components discovered by several cybersecurity firms and government agencies. The LazyCargo malware is a Windows dropper for another module in the framework. I don&amp;rsquo;t have access to any of the other components, so I wrote a payload to run the LazyCargo malware at the end of the analysis to verify my findings from the static analysis.&lt;/p&gt;</description></item><item><title>About</title><link>https://blog.lukasmay.com/about/</link><pubDate>Sun, 07 Dec 2025 18:55:14 -0500</pubDate><guid>https://blog.lukasmay.com/about/</guid><description>&lt;h1 id="about-lukas-may"&gt;About Lukas May&lt;/h1&gt;
&lt;p&gt;Welcome to my cybersecurity blog! I&amp;rsquo;m a Junior at the Rochester Insitute of Technology (RIT) studying cybersecurity who loves diving deep into interesting technical problems and sharing what I learn along the way.&lt;/p&gt;
&lt;p&gt;A big part of my focus is OT security&amp;ndash;how we defend critical infrastructure like the electric grid, water treatment plants, or manufacturing systems. Through this site I aim to share my research and analysis of the cybersecurity and tech world.&lt;/p&gt;</description></item></channel></rss>