Machine Learning engineer
Kaggle competitions master
Having dedicated 6 years to machine learning and data science, I possess a wealth of experience in developing quantitative models across diverse domains. My expertise, cultivated through continuous learning and active engagement in Kaggle competitions, spans various facets of the field. I am adept at leveraging cutting-edge technologies to tackle complex challenges, driven by a passion for innovation and problem-solving. Throughout my journey, I have honed my skills in data analysis, modeling, and interpretation, equipping me to deliver impactful solutions that drive business success and advance research in the ever-evolving landscape of machine learning and data science.
Kaggle Achivement
As a Kaggle competition master, I relish the opportunity to tackle new problems and push the boundaries of my expertise in machine learning and data science.
Google QUEST QA Labeling
- 4th Place/1571 Teams, Gold Medal
- Implemented a cutting-edge 3-head BERT model to discern and learn the intricate interaction patterns among the various inputs in the competition. This approach contributed to the development of a robust model, enhancing the overall performance in QA labeling
RSNA-MICCAI Brain Tumor Radiogenomic Classification a
- 1st Place/1555 Teams, Solo Gold Medal
- Implemented a sophisticated validation technique to mitigate overfitting challenges, particularly crucial given the limited size of the competition dataset.
Prostate cANcer graDe Assessment (PANDA) Challenge
- 15th Place/1010 Teams, Silver Medal
- Implemented a data centered strategy by applying heavy augmentation techniques on the extracted tile cutouts. Additionally, employed a variety of different Convolutional Neural Networks (CNNs) to enhance the model’s ability to discern intricate patterns in prostate cancer images.
YOUTUBE VIDEOS
I find great joy in sharing my learnings with the community, actively engaging in the Tunisian AI community to provide assistance and support to individuals striving to achieve their goals in the field of artificial intelligence.
OTHER ACHIVEMENT
Data Science Game Finalist - Paris
Our team was the first and only from the African and MENA region to reach the final of the most prestigious inter-university Data Science Game worldwide, solving industry-led challenges among the top 20 global teams at Château Les Fontaines, Paris.
Learn More
Solo Winner - UmojaHack:
UmojaHack Africa, the continent's premier hackathon, offers $10,000 in prizes, engaging over 2,000 students from 300+ universities across 30 countries. Transitioning to a virtual format in 2020 and 2021, the 2022 event, enhanced by Zindi, aimed to broaden participation and amplify the collaborative experience, supported by major tech sponsors. I applied computer vision techniques to classify images of deep-sea invertebrates, focusing on species found off the coast of South Africa. Leveraging advanced algorithms and post-processing methods, I achieved an impressive accuracy rate of nearly 99%, showcasing the efficacy of the developed model in species classification.
Learn More
Solo Winner of the Electricity and Gas challenge
The biggest AI Hackathon in Africa and the Middle East (35+ countries). Organized by InstaDeep, Google, and the Ministry of Industry, Mines and Energy, the event brings together students, developers, tech entrepreneurs, researchers, Ph.D. holders, and startups to create innovative solutions using Artificial Intelligence.
Learn More
Runner-up in a INFOR Hackathon
In this competition, participants will dive into a rich dataset to forecast future hotel prices based on historical booking and pricing data. The challenge revolves around understanding and predicting hotel prices using various data fields that include booking numbers, revenues, hotel prices, and competitor prices across different time lags.
Learn MoreGithub Projects
RAG Responder
RAG Responder is a tool that utilizes Retrieval-Augmented Generation (RAG) techniques to answer questions related to a document. Currently, it supports documents with .txt file extensions. This repository provides a step-by-step guide on how to set up and use the RAG Responder.
Learn More
Supersize-GAN
Supersize-GAN is a fun project inspired by the SRGAN paper, implementing a modified version of the SRGAN architecture. The primary goal of this project is to provide a simple and fun implementation while refreshing my legacy knowledge in GANs.
Learn More
RSNA-MICCAI Brain Tumor Radiogenomic Classification
Achieved first place in a Kaggle competition with a simplified 3D CNN model, overcoming challenges in model validation and data handling through an innovative approach to imaging, demonstrating that simplicity and a novel technique can outperform complex solutions.
Learn More
Fraud Detection in Electricity and Gas Consumption Challenge
Developed a model for Tunisia's STEG to detect fraud in electricity and gas consumption by analyzing clients' billing histories, aiming to reduce the company's financial losses, estimated at 200 million Tunisian Dinars.
Learn More
AI4D Takwimu Lab - Machine Translation Challenge
This repository contains the 3rd place solution for the AI4D Takwimu Lab - Machine Translation Challenge competition. The challenge focused on translating text from French into Fongbe or Ewe, two Niger–Congo languages that are part of the Gbe cluster.
Learn More
Early Detection of Diabetic Retinopathy
This repository houses a machine learning model for detecting diabetic retinopathy, supporting Aravind Eye Hospital's blindness prevention in rural India. It aims to automate diagnosis, enhance screening, and potentially identify other eye diseases, improving healthcare access and outcomes in underserved communities.
Learn MoreLet's Connect and Innovate Together!
Reach out for collaboration, questions, or just a friendly chat. I'm here to discuss machine learning, data science, and everything in between. Click below to drop me a message and let's embark on exciting possibilities together.