KINGUE

Flore is an ultra-low latency trading tool operating on parallel markets such as Polymarket. I used C++ to meet the performance and latency demands, and it's currently running on AWS bare-metal instances. I use two algorithmic processes to trigger orders, with sub-20 µs reaction time. This project is in constant progress and improvement; the latest live metrics are at the bottom of this page.

The Challenges

To build this, I've been consistently solving new challenges along the way. Here's the full story from start to current.

When I went on the BTC 15-minute market, I noticed something strange that caught my attention:

So after some serious grinding, I finally found different live feeds that gave me a clear edge over the streams that other arbitrage bots and Polymarket were using. With that, I started building the main algorithm:

The algorithm was a tough build. The stream was really raw and inconsistent, full of spikes and random noise. I had to clean everything out myself.

The algorithm was okay, but I wanted something that could evolve with time. Thankfully, I had some machine learning experience. That's when I started recording hundreds of markets, built my own dataset, and trained my own MLP.

So I went back to my dataset and ran a bunch of data visualizations to understand what was happening, and that's when it clicked!

I fixed my dataset, trained it again, and it worked! Here are the backtesting results that confirmed it:

Here is a live presentation of my MLP running on 105 unseen markets, around 25 hours of recorded data.

Unfortunately, despite all that work, two big problems were still standing in the way.

1. My backtests were not faithful to reality. Polymarket's latency is an unknown variable, and nobody has a way to model it. Every latency model I built fell apart: from one moment to the next, an order would go from 150 ms to 450 ms, and in those 300 ms the price could double while the market makers pulled their orders.

2. Polymarket itself was my biggest problem. Beyond latency, every two weeks a new trading rule came in, changes that completely reshuffled the way market makers priced the shares. Each time, that meant recording the markets all over again and retraining a model that had never even seen production.

Because of that, I took a few weeks off to think, and went back to my YouTube community, where I teach people how to develop in C++. Then one evening I decided to start again, this time integrating AI far deeper into the way I create algorithms.

And that's when it clicked.

Between July 24 and September 8, I ran 9 algorithms live:

And that's where I'm at now:

In the winter of 2024, I led a community around low market cap tokens, gathering over 800 members to discuss trading strategies, build projects together, and promote my own tools: copy trading bots, sniping programs, and Reflexion, a stealth bundler designed to influence low-cap token charts. Combined, these tools ran over 500K in monthly volume.

The Story

I started getting interested in lowcap tokens when a friend of mine showed me how volatile those markets were. I soon joined a community myself, but the people running it were making poor decisions for the group and taking the community in the wrong direction. So I decided to build my own, quickly outgrowing my original base.

I used this community to launch my own coins and promote them. They would often reach over 50K market cap, sometimes going as high as 80K.

This is where I saw the real potential. I first started building a sniper of my own that would get me quick wins I could reinvest into designing and building my platform:

The goal was simple: share my sniper with people and take a percentage of the volume.

I soon built out a whole platform with copy trading, where traders and users could sign in and use my services. Over 500K in volume per month with just copy trading and sniping.

But this was only the start. The biggest part of Deckflow were the private clients.

I built a tool that creates fake movement on a token's chart and could invite people to invest, although the patterns were completely synthetic and the whole liquidity was owned by a single person. I called that tool:

What you just saw was a perfect synthetic hype on a token, built only by perfectly timing transactions on the blockchain. To be honest, this is pretty impressive, all of that in C++.

Contract: AsW1dqhhxp9GuTKfWREPh6dkp7JiVFshpFxqAv4Jpump

This was only proposed to clients with high win potential.

Growing up, I spent countless hours playing Valorant, and like many players, I quickly noticed how Riot Games was heavily criticized for the inflated prices of their in-game skins and digital assets.

That's where the idea came from: build a marketplace where people could buy and sell accounts with skins and in-game assets at a fraction of Riot's prices, a real alternative to their official store. That's how Euwshop was born.

We grew incredibly fast, over 3,000 users in the first two weeks alone.

That's when we saw the real potential of scaling into a worldwide marketplace.

We also invested heavily in communications, especially on TikTok. Our community manager Zartex generated over 1 million views promoting Euwshop.

We expanded the website with in-depth data visualization, applied proven sales techniques, and completely redesigned the platform to improve conversion rates and drive purchases from potential clients.

In the end, Euwshop was a real success, both for me and for my team.

This came to mind after going on Vinted and watching people reselling shoes at crazy prices, while on the official manufacturer's website they were way less expensive.

The demand was so high that pairs would sell out in minutes, and if you wanted them after that, you'd have to pay up to 3x the original price.

So I started developing an algorithm that, during shoe drops, could buy pairs faster than any human, in bulk, to resell them on our own platform.

That's how Bothlow was born.

We first launched a waitlist to evaluate the market and quickly found ourselves with over 800 sign-ups. Thanks to that, we were able to pitch our friends & family and cover the basic costs in exchange for 10% of the company.

That's how we opened our site.

And to promote the business, we also started investing in professional shootings.

I learned what it truly means to co-found something, to work with a partner, a friend. How to divide tasks and settle disagreements with maturity, always keeping the company at the heart of the work.

In the end, this was my first hands-on business experience. Meeting with legal advisors, reading and complying with all the economic laws around company creation in France. I came out of it grown, and full of ambition for what was next.