How I Got Started
My programming journey began with modifying Xbox 360s for game cheating in middle school. What started as using other people's tools quickly evolved into creating my own when I realized how expensive it was to stay undetected online - stealth services cost $20/day and key vaults were another $20 each. There's a my 2014 Xbox modification video from when I got my Xbox modified in 2014.
I taught myself C# and C++ by reverse engineering games and building cheat tools, I even learned how to install the chips into Xbox 360's, some photos can be seen my Instagram build photos. Eventually, I was creating and selling my own cheats, which can be seen my original cheat showcase video. When competitors cracked my software, I got creative - I released an intentionally inferior "free" version that made my paid version look better by comparison.
High school introduced me to Java through a programming class, which felt familiar after C#. A teacher's recommendation led to my first real job offer - a computer technician role with the promise of becoming a developer in 3-5 years. I was promoted to Junior Programmer after just one year.
During my free time, I pivoted to automation - building tools for online checkouts and restock monitoring. The skills from game hacking transferred perfectly, especially around detection evasion. When I couldn't bypass a particular user-agent check, I injected a payload directly into Chrome to hook the function returning the user-agent. Apparently nobody had done before I guess, someone complimented the idea, calling it a "novel idea."
Today I'm full-stack by choice, though I deliberately avoid heavy frameworks like React. I believe in performance over convenience - why download 500KB of JavaScript when you can build the same functionality more efficiently without it?
An Xbox 360 cheat for multiple Call of Duty games, aimbot, esp.
A cheat for Call of Duty Black Ops 3 on PC, including aimbot, no spread, and ESP.
A cheat for CSGO on PC, including aimbot and ESP.
This entire project has a lot of personal significance because there's a lot of stuff that I did that I was never planning on doing, all because I wanted a specific truck -- and I eventually got the truck. The truck is a Chevy Silverado Regular Cab Short Bed (2003-2007), and I looked for a few months and couldn't find one for a reasonable price.
Monitoring IAAI
I created monitors to watch websites like IAAI for new posts and send Discord webhooks so I could get notifications right away, never really found anything worth paying to get shipped home.
Scraping Facebook Marketplace
Then I created a Facebook marketplace scraper.
Scraping multiple zip codes around the US to cover as far as I was willing to go, because there was a maximum radius at one point.
I noticed a lot of extended cabs that were incorrectly labeled as "two doors", and that got really annoying.
(Facebook got locked)
Building a dataset
I then started to look into training yolo models for custom object detection. I learned I would need a lot pictures of the type of truck, and basically anything else. So I started scraping car sales websites.
Labeling the data
I made a simple C# application to make labeling the data quicker. This made it where I could just draw boxes around the bounds of the vehicles, and resize the image and save the scaled bounds before moving to the next image. Did that for around a thousand images.
Data labeled
Training the model
I started training, using the GPU in my computer, and let it run overnight and see what happened. Checked back and it was detecting things like cars and tractors, which were kind of close, but it wasn't detecting houses.
Refining the model
So I scraped and labeled some more data, focusing on getting good, diverse samples. I started the training process one more time and just let it run overnight again.
Model worked
When I checked it, it was actually working, correctly identifying the regular cab short bed trucks I was looking for.
It was smart enough to still detect them even if the truck had a topper on the back.
Detection showing successful label highlight on clean white Silverado
Waiting
About a month later, I got a notification, my scraper found a truck, seven hours away in the Lake of the Ozarks.
Truck found
We left the next day, drove straight there, slept for a few hours and then met the guy in the morning. Went for a test drive, bought the truck, and drove all the way back home.
Wouldn't be a Chevy without some rust. I cut it out, learned learned to weld and welded in patch panels.
A machine learning model trained to generate human-like cursor movement between two points. Trained using TensorFlow, using a dataset I built of my own mouse movement.
Clipt AI, a comprehensive video editing platform built with vanilla HTML, CSS, and JavaScript for the frontend and a C# backend, designed to streamline content creation by using AI to automatically generate engaging short clips from long-form videos.
A full-stack web application built with vanilla HTML, CSS, and JavaScript, powered by a C# backend and MySQL database, that provides a sophisticated AI development suite for creating, managing, and interacting with custom LLM agents through a standardized API and an agent messaging system.
This is a comprehensive Amazon seller tool, built with HTML, CSS, and JavaScript for the frontend and a C# backend, designed to provide in-depth product analysis, profitability calculation, and competitive insights by leveraging data from the Keepa API.