Resume

I'm Abdenour, a Machine Learning Engineer. This is my hand-crafted portfolio.

Technologies

Python
PyTorch
Keras
Scikit-Learn
TensorFlow
Numpy
Matplotlib
Pandas
Jupyter
git
GitHub
CI / CD
SQL
uv
Ruff

Main experience

  1. Machine Learning Engineer (work/study training)

    Sept 2023  -  Sept 2024

    ST Microelectronics

    • Developed a variational autoencoder (VAE) with PyTorch for dimensionality reduction and feature extraction from high-dimensional simulation data.

    • Accelerated electronics simulations, cutting runtime from 25h to under 3h while maintaining <3% error, by leveraging Deep Neural Networks, Multi-Regression models, and parallelization (PyTorch, Scikit-Learn, Keras, Joblib).

    • Automated hyperparameter optimization with Optuna, improved experiment traceability and reproducibility with MLflow.

    • Built a 5GB structured dataset from 400+ circuits across various voltage/temperature simulations, with Pandas and SQL, using a custom data collection pipeline to enable large-scale modeling and analysis.

    • Enhanced research team collaboration by developing and deploying comprehensive technical documentation (Sphinx, LaTeX) covering both successful and unsuccessful algorithms used throughout the project.

    Python
    PyTorch
    Scikit-Learn
    Keras
    Optuna
    MLflow
    Joblib
    Pandas
    SQL
    Sphinx
    LaTeX

  2. Machine Learning Engineer (intern)

    May 2023  -  Sept 2023

    ST Microelectronics

    • Developed a Deep Reinforcement Learning prototype to compare with traditional optimization algorithms in high-dimensional search spaces with Actor-Critic & PPO, using Stable-Baselines3, PyTorch, and OpenAI Gym.

    • Reviewed research literature on state-of-the-art Reinforcement Learning methods used in the electronics industry.

    • Designed an interactive dashboard to visualize agent trajectories on dimension-reduced data (UMAP, t-SNE, Kernel PCA), enhancing interpretability of both Reinforcement Learning and optimization algorithms.

    Python
    OpenAI Gym
    Gymnasium
    Stable Baselines 3
    Scikit-Learn
    PyTorch
    Matplotlib
    Pandas
    LaTeX

  3. Open Source Developer

    2023  -  current

    Python & AI Ecosystem

    • Scikit-Learn: Contributed a new evaluation metric (macro averaged MSE) for ordinal classification, including full documentation and testing (scikit- learn-contrib/imbalanced-learn)

    Python
    Scikit-Learn
    git
    GitHub
    CI / CD
    Pytest
    Sphinx

Other experience

    Clovis

  1. Technological startup project: “Projet Clovis” (link)

    2018 - current
    Clovis
    Fig. Document creation, with block-style editor: automatic styles to focus on content only.

    • Built & scaled a full-stack EdTech platform (mobile & web) enabling students to create flashcards and study through interactive mini-games.

    • Grew the platform to 350+ active users, with 40,000+ flashcards and 2,200+ study sheets created.

    • Pitched the project in front of 100+ people at major events (Le Village by CA, digiSchool HYPE Awards 2018), with a featured video on YouTube.

    Fig. One of the mini-games to study your flashcards: choose the right answer in limited time.

    • Led development over 7+ years, from inception at 17 years old to a fully maintained large-scale project.

    • Tech stack: Python (Django), SQL, TypeScript, React, PHP.

    • Managed databases & analytics: phpMyAdmin, Google Analytics for user insights.

    TypeScript
    React
    Tailwind CSS
    Python
    Django
    MySQL
    PHP
  2. Python Prépa

    Python/Algorithms Teacher (students & YouTube) (link)

    2020 - 2021

    • Launched an educational YouTube channel on the subject, reaching 19K views and 1,300+ hours of watch time

    • Taught intensive private lessons for several months to top-tier science foundation students (PCSI, PC, BCPST)

    • Designed personalized learning programs and created comprehensive course materials, including exercises and detailed corrections

    Python
    YouTube
    LaTeX

Daniel lewis

Richard was hired to create a corporate identity. We were very pleased with the work done. She has a lot of experience and is very concerned about the needs of client. Lorem ipsum dolor sit amet, ullamcous cididt consectetur adipiscing elit, seds do et eiusmod tempor incididunt ut laborels dolore magnarels alia.

Projects

A list of my projects.

Highlighted projects

Recognize Handwritten Digits & Letters (link)

Fig. Live demo recognizing digits in real time.

• Built a real-time handwritten letter & digit recognition app achieving 97% F1-score on 26 uppercase and lowercase letters (+ digits)

• Designed a lightweight CNN model, optimized for web performance and deployed in production

• Trained on 700MB of data using PyTorch and ONNX, with instant client-side inference with ONNX Runtime, React & Tailwind CSS

Python
PyTorch
ONNX
React
Tailwind CSS
Scikit-Learn
Jupyter

Competed in 8 Kaggle AI Competitions (link)

Fig. My Kaggle leaderboard, sorted by best results. I ranked 3 times among the top 33%.

Best ranking 105/934, using XGBoost, CatBoost, LightGBM, & Random Forests

Python
Kaggle
XGBoost
CatBoost
LightGBM
Scikit-Learn
Pandas
Matplotlib
Jupyter

AI & Data projects

Stock Analysis (link)

Fig. Apple stock "close" prediction compared with actual value, done with Prophet model (from Meta).

Stock prices analysis & prediction using time series methods.
I benchmarked Prophet, XGBoost & SARIMA models, and visualized the data on interactive dashboards.

Python
Prophet (Meta)
XGBoost
Scikit-Learn
Pandas
Matplotlib
Jupyter

Customer Segmentation (link)

Fig. Cluster results: I segmented the client base into 3 clusters & found concrete criteria to identify the most valuable clients (here, they are mostly highly educated and purchase in real stores more than online).

Customer segmentation & marketing campaign success prediction.

Python
Scikit-Learn
Keras
TensorFlow
Statsmodels
Pandas
Matplotlib
Jupyter

Other projects

Orientate Undirected Graph (link)

Fig. Excerpt from the LaTeX report

The objective of this project is, given an undirected graph, to transform it into a strongly connected graph using Schmidt's chain decomposition.

Python
NetworkX
Matplotlib
Jupyter
LaTeX

My Own Compiler (link)

Fig. Code snippet of the language

I created my own programming language with my own compiler, using ANTLR4.
Current state of the project: currently rewriting from Java to Python, with a target to LLVM.

Java
ANTLR4
LLVM
Python

Epidemic Simulation (link)

Fig. Evolution of p-values for Kolmogorov-Smirnov statistical test.

Epidemic simulation with Galton-Watson model, in order to empirically prove the Yaglom theorem.

Python
Pandas
Matplotlib
Jupyter
LaTeX

Education

My studies.

Education

  1. MSc Applied Mathematics and Statistics

    2022 - 2024

    Université Clermont Auvergne

    - Machine Learning | Deep Learning
    - Data mining | text mining
    - Time series forecasting
    - Bayesian statistics
    - Stochastic algorithm | Markov chain | genetic algorithm
    - Linear optimization | mathematical modeling | simplex algorithm
    - Database | SQL
    - Introduction to financial mathematics
    - Survival analysis | failure prediction

  2. BSc Computer Science

    2019 - 2022

    Université Clermont Auvergne

    - Collaborative projects | Git | GitHub Actions | code reviews
    - Text recognition | OpenCV | Python
    - Scientific libraries | Numpy | Matplotlib | Pandas | SymPy
    - Linux | Command Line Interface | Docker (intro)
    - Mathematical finance basics | discounting | compound interest | IRR
    - Algorithmics | dynamic programming | graph theory | C++ | performance testing
    - Formal languages | regular expressions | automata | language creation | ANTLR4 | Java