About Me

My name is Fares Ben Slimane, and I am a Computer Vision and AI enthusiast. I graduated with a Master’s degree in Computer Science, specializing in Artificial Intelligence, from the University of Quebec at Montreal (UQAM). I currently work as a Machine Learning Software Developer and a part-time tutor. I have a strong background in AI research and development, with expertise in machine learning, speech recognition, and computer vision. My passion lies in leveraging technology to make a positive impact on the world.

Here are a few highlights from my professional journey:

AI Mentor at OpenClassroom: I instruct advanced AI courses to professionals, covering topics such as data analysis, classical ML, NLP, CV, and AI project management.
Machine Learning Developer at Fluent.ai: I contributed to developing and integrating AI models, reducing model sizes and improving performance in multi-wakeword scenarios.
R&D Machine Learning Developer at Hummingbirds AI: I implemented state-of-the-art computer vision algorithms and led research initiatives to address real-world challenges.

Download my full CV

Technical Skills

Programming Languages: Python, C++, Javascript, Dart(Flutter)
Libraries & Tools: NumPy, Pandas, Scikit-learn, OpenCV, TensorRT, ONNX, Git, AWS, GCP, Asure
Deep Learning Frameworks: TensorFlow, Keras, PyTorch (preferred)
Machine learning: Data analysis & visualization (Matplotlib, Plotly..etc), Supervised classification and regression ML algorithms, unsupervised ML algorithms, Statistics, Probability, Calculus and Linear algebra.
MLops & Cloud: Git, Docker, Kubernetes, AWS, Google Cloud Platform, Azure ML, MLflow, ClearML
Libraries & Tools: NumPy, Pandas, Scikit-learn, OpenCV, TensorRT, ONNX unsupervised ML algorithms, Advanced statistics, Probability, Advanced Calculus, Linear algebra and optimization
Deep learning (Theoretical & Practical): Neural Networks & Convolutional neural networks (CNNs), Recurrent Networks (RNN, GRU and LSTM), Transformers (attention-based models), Generative Models (GAN, VAE), Reinforcement Learning (Q-learning, Monte Carlo, Sarsa, PPO..etc) and LSTM), Generative Models (GAN, VAE), Reinforcement Learning (Q-learning, Sarsa, PPO..etc)
Mila Course (Prof. Aaron Courville) - Representation Learning (IFT 6135) Winter-2019, Grade: A-
Computer Vision: Image classification and segmentation, Object and anomaly detection, Video and Image Analysis
Speech Recognition: Lightweight Wakeword Detection, Command detection, ASR, Speech-to-Text and Text-to-Speech
Natural Language Processing:Text classification, Language modelling, Machine translation, LLMs
Soft Skills: Worked in high-paced startup environments, Led R&D ML projects, Team Player with efficient communication skills, Positive force and a good motivator, Effective in both spoken and written English and French

Projects

Sign Language Recognition & Translation (2019) Build a system that interprets a sequence of images, representing sign language, and generates a coherent textual translation in spoken language.
GitHub project Link
Sign Language Tutoring System An automated system that teaches sign language to non-deaf users, featuring real-time gesture recognition and an ergonomic Human-Machine Interaction Interface.
GitHub project Link
Using AI for Automated Hydroponic Growing System A smart hydroponic system that uses AI to control indoor settings and detect plant anomalies and diseases.
Tracking and Predicting Student Performance A predictive model that tracks students’ academic performance and forecasts future success using data from UQAM’s SPARI.

Publications and Talks

Conference Paper: Context Matters: Self-Attention for Sign Language Recognition (ICPR 2020) Arxiv Link

Hands on RL algorithms
By Fares Ben Slimane
Posted on July 11, 2024

This notebook is a tutorial to explain and showcase how to use RL algorithms like Q-learning (model-free version and the DQN version), Sarsa, MC, and how they differ, using PyTorch and the OpenAI Gymnasium library. This notebook will give you a straightforward overview of how RL algorithms work with real... [Read More]
Tags:
OpenAI Hands-on

People are usually familiar with supervised or unsupervised learning but not Reinforcement Learning. In RL, we choose an action using a policy and given some data to maximize the expected long-term reward (all these terms will be explained).
By Fares Ben Slimane
Posted on July 9, 2024

Context and Plan [Read More]
Tags:
Introduction to Reinforcement Learning

People are usually familiar with supervised or unsupervised learning but not Reinforcement Learning. In RL, we choose an action using a policy and given some data to maximize the expected long-term reward (all these terms will be explained).
By Fares Ben Slimane
Posted on July 9, 2024

Disclaimer (You can ignore this part and start with the next section.) Why am I writing these blogs? Especially since, as my brother told me the other day, there’s a good chance that no one is going to read this. I am a strong believer in the Feynman technique. I... [Read More]
Tags:
Teaching a model how to count with Logistic regression

In this quick post, we will learn how to build a Logistic regression model from scratch that learns how to count digits. The idea is that the model will learn how to count from 0 to 9 in binary.
By Fares Ben Slimane
Posted on January 2, 2017

Requirements : [Read More]
Tags:

About Me

Technical Skills

Projects

Publications and Talks

Hands on RL algorithms

OpenAI Hands-on

People are usually familiar with supervised or unsupervised learning but not Reinforcement Learning. In RL, we choose an action using a policy and given some data to maximize the expected long-term reward (all these terms will be explained).

Introduction to Reinforcement Learning

People are usually familiar with supervised or unsupervised learning but not Reinforcement Learning. In RL, we choose an action using a policy and given some data to maximize the expected long-term reward (all these terms will be explained).

Teaching a model how to count with Logistic regression

In this quick post, we will learn how to build a Logistic regression model from scratch that learns how to count digits. The idea is that the model will learn how to count from 0 to 9 in binary.