[nom]

Rabat, Morocco

# [e-mail] · I [téléphone] [nom] · [nom]

Education

AI Engineering · Ensias, Rabat

2023 - 2026 · Relevant coursework: Machine Learning, MLops, Data Preprocessing

MP* Classes · AL ZAHRAOUI Center, Rabat

2022 - 2023 · Relevant coursework: Advanced mathematics and physics

MPSI Classes · MOHAMMED VI Center, Kenitra

2021 - 2022 · Relevant coursework: Mathematics, Physics, and Engineering Science

Professional Experience

AI Engineer Intern – Sofrecom, hybrid · March 2026 - Now

• Built a production RAG pipeline combining a vision LLM microservice (structured JSON extraction from PPTX/PDF) with hybrid search over a vector database for automated marketing brief analysis.
• Engineered a multi-service architecture (FastAPI 3 ports) orchestrated via Docker Compose, with parallel batch processing, retry/fallback logic, and a JSON Schema contract enforcing strict data handshakes between services.
• Implemented a scalable data ingestion pipeline featuring SQLite embedding caching, checkpoint/resume on crash, dead-letter queuing, and hybrid retrieval (dense 2 + sparse BM25) over a Qdrant vector store.
• Delivered a dual CLI + REST API interface with Pydantic-validated I/O at every boundary, supporting both local and fully containerized deployments
• Skills: Vision LLMs, prompt engineering, RAG, embeddings, LangChain, Qdrant, FastAPI, Docker, SQLite caching Data Scientist Intern - Smart Automation Technologies, Remote · August 2025
• Developed and optimized a Hidden Markov Model (HMM) based fault detection system, achieving 94.12 % accuracy in identifying boiler plant anomalies.
• Engineered a comprehensive data preprocessing pipeline for time-series data, integrating techniques for missing values, heavy-tailed distributions, and feature engineering to prepare diverse datasets for HMM training.
• Implemented a specialized HMM architecture with global and class-specific fine-tuning, leveraging Gaussian Mixture Models for emission probabilities to accurately classify complex sequential fault patterns.
• Skills: Python, Scikit-learn, Pandas, NumPy, Hidden Markov Models (HMMs), Anomaly Detection, Data Preprocessing, Pipeline Development.

Projects

Mushakkil: An Arabic Text Diacritization system.

• Curated and preprocessed a subset of the Tashkeela corpus dataset for model training.
• Implemented and compared multiple models: HMMs, RNNs, LSTMs.
• Tools: Python, PyTorch

Implementing a Federated Learning Algorithm

• Studied core principles and challenges of federated learning, including data heterogeneity and communication constraints
• Implemented the FedHe algorithm (from the paper FedHe: Heterogeneous Models and Communication-Efficient Feder- ated Learn- ing) in a simulated federated environment
• Tools: Python, PyTorch

Key Points Detection and Tracking for Event Detection in Crowded Environments

• Developed a system to detect and track key points in video frames using Harris Corner Detection and Lucas-Kanade Optical Flow.
• Implemented magnitude and directional modeling of motion vectors using Gaussian Mixture Models and Von Mises Mixture Models
• Designed a block-based grouping strategy and a tracking system to maintain group identities across frames, enabling effective event analysis in crowd scenes.
• Tools: Python, NumPy, Scikit-learn, Gaussian Mixture Models, Von Mises Distribution Computer vision.

Maritime Anomaly Detection with AIS Data

• Developed a pipeline for analyzing AIS maritime data to detect anomalies and predict vessel types and statuses.
• Built an end-to-end system including, preprocessing, feature engineering (PCA, mutual information), and Random Forest modeling.
• Addressed major data challenges including missing values, outliers, and class imbalance; integrated both dynamic and static AIS data.
• Skills: Python, Scikit-learn, Pandas, SMOTE, ADASYN, PCA, Random Forest.

Skills

• Programming Languages: Python, SQL, Bash
• Frameworks & Libraries: PyTorch, Scikit-learn, LangChain, FastAPI
• AI & LLM: Vision LLMs, Prompt Engineering, RAG, Embeddings (BGE-M3)
• Vector & Data Tools: Pandas, PostgreSQL, MongoDB, PySpark, Qdrant, SQLite
• DevOps & Tools: Docker, Git, GitHub
• Spoken Languages: Arabic, French, English