Aziz Belaweid's Projects
Instruct-tune LLaMA on consumer hardware
📚 Papers and blogs by organizations sharing their work on data science & machine learning in production.
Google Brain AutoML
My portfolio repo
Data-Colab is a Tunisian NGO that aims to upgrade and help AI community in Tunisia, this is my work on their test to join their engineering department fortunately i was accepted. The test consists of 3 parts the first part is about computer vision and transfer learning the second part is about NLP and the third part is about general datascience and artificial intelligence.
My work on the global terrorism database in Kaggle, my goal was to explore different machine learning algorithms, their validations and test out diffeent scoring metrics and perform EDA o n the Dataset.
My work during EPT's mini AI hackathon the goal was to create a regression model to help hotel pricing decision making using GridSearchCV along woth CatBoostRegressor I managed to get 4'th place. The data can be found on kaggle : https://www.kaggle.com/c/ai-mini-hackathon-ept/data
I wanted to create my own FER program that works real time to further expand my knowledge on CNN, Data augmentation and Transfer Learning. Data is used from a kaggle competition, models architectures are purely my own using Keras I have achieved 64% accuracy which I think is decent but there's space for improvement. I didn't work on face detection myself but my future goals are to implement face detection myself, work on real time recognition and Improve model's performance.
This repo contains my code for learrning FastAPI, CRUD, Async operations
This is my work for AI HACK qualification, my goal was to explore as many classification models as i can, i tried some feature engineering techniques and modified multiple featues. The models I used are KNN, Random Forest, Decision Tree, MLP, AdaBoost, XGBoost I used ROC/AUC to compare between models and accuracy aswell finally I chose the best models and applied Stacking to them which gave me the best result. I explored aswell other techniques such as PCA, LDA and SMOTE because the data was unbalanced, I also built a small NN using Keras. The data can be found on Zindi : https://zindi.africa/competitions/financial-inclusion-in-africa/data
Distributing Flan Model using Jina Ecosystem
Google IT Automation with Python Professional Certificate - Practice files
Cloud-native neural search framework for 𝙖𝙣𝙮 kind of data
Implementation of binary and categorical/multiclass focal loss using Keras with TensorFlow backend
Models and examples built with TensorFlow
Objectron is a dataset of short, object-centric video clips. In addition, the videos also contain AR session metadata including camera poses, sparse point-clouds and planes. In each video, the camera moves around and above the object and captures it from different views. Each object is annotated with a 3D bounding box. The 3D bounding box describes the object’s position, orientation, and dimensions. The dataset contains about 15K annotated video clips and 4M annotated images in the following categories: bikes, books, bottles, cameras, cereal boxes, chairs, cups, laptops, and shoes
🤗 PEFT: State-of-the-art Parameter-Efficient Fine-Tuning.
Tensors and Dynamic neural networks in Python with strong GPU acceleration
Sentiment Analysis task on Tunisian and Arabic dialect, data augmentation for NLP and scrapping google maps for more data
This is my work on the kaggle comeptition Disaster or Not. It's about classifying tweets : tweets that are reporting actual and real disasters and tweets that aren't. In my work I used a lot of NLP techniques word2vec / TF-IDF / lemmitazation / Words clouds. In modeling I applied ensembling and boosting models to get best Results. Next up I ll try embeddings and BERT.