I build a GPT3-like Large Language model based on the company I work for's data. With more than 50 M parameters trained on high-performance computer I got amazing results. The model was capable of mimicking the way of writing maintenance reports of offshore well. This kind of report is very important and very jargon-based. The level of expertise to write those reports is very high, and a similar LLM could help to reduce the needed experience to be able to write a report.
This project is usually the first step in advanced AI programming. We aim to make an AI able to recognize which animal is in a picture. I used a convolutional neural network architecture. Thanks to TensorFlow I easily used convolutional neural networks, TensorFlow is a powerful library for AI programming/
Example of a fail.
We developed an RPG game on the MVC model using Qt framework. The Game was developed using Qt 5.15 and opencv4
The source code is free on Git Hub.
Several log analysis of a date frame(~60GO) to train a model to predict attack
The source code is free on GitHub.
Ask me if you want a very complete report
During my 4th year I did a 3 months internship, my goal was to build a website and a database for Human Resources Director. They can: put into a database resumes and CVs of interesting people, put an automatic reminder to remind them to contact a candidate. They also have access to the database for statistical purposes.
The source code is free on GitHub.
User interface to add a CV to the database.
Admin interface.
This project was a 5 students project, I focused on building an AI able to understand what is the best pressure to grab an item without breaking it. I created an algorithm inspired by Darwinian evolution theory. Every time a test is fail the algorithm modifies the maximum or minimum pressure on the item.
A quick example of how genetic algorithm works.
During my scholarship I trained my programming skill in basic games like Connect 4 or Zelda.
Interface of a game, white versus red.
Virtualization and Software-Defined Networking (SDN) technologies are gradually gaining ground not only in data centers but also in WAN and RAN due to several key features mainly flexibility in scaling in and out the allocated resources and simplicity of configuration and troubleshooting. SDN applied to WAN enables management in a centralized manner and thus provides the ability to enhance application performance thanks to the global on-time knowledge of the network resources utilization and the demande requirement.
Our hand made poster.
