Six selected projects · an archive
Work
Chosen for technical difficulty, my personal share and the quality of the evidence. Every status is shown as it is: prototype, research, live product or academic project.
TotalEnergies OneTech · DataLab, Pau
FlagshipThe right value, for the right object: LLM extraction from industrial reports
Two-pass extraction guided by document structure, on self-hosted LLMs and GPUs, with an evaluation that separates identifying the objects from the quality of their attributes.
Evaluated prototype, candidate for integration
Independent product · real time, multi-instance
FlagshipMaison Dlala: live auctions where the server is always right
A four-language auction marketplace, Arabic included: idempotent bids inside one atomic write, a server-owned clock, live rooms spread across several instances.
Live — maisondlala.com
CETIC · NLP, GenAI and cybersecurity
FlagshipRiskTailor: one risk report, one rendition per reader
A RAG pipeline behind an API that tailors a cyber-risk report summary to the architect, the manager or the compliance lead — published at INFORSID 2026, as first author.
Published research — INFORSID 2026
MSc DataScale · joint project with Youcef Bouagada
Rules and an LLM that check each other: extracting a verifiable knowledge graph
French scientific text → RDF/OWL triples: NER, syntactic rules, LLM fallback, ontology validation and OWL-RL reasoning. F1 0.81 against 0.73 (rules) and 0.75 (LLM alone).
Academic project — 50-sentence corpus
MSc DataScale · data integration and quality
Measuring quality before promising it: 38 indicators on energy data
Consumption, population and territorial reference data reconciled by IRIS zone: daily Airflow orchestration, Talend and Apache Hop transforms, 38 quality metrics tracked before and after cleaning.
Academic project
Data engineering project · simulated scenario
An open-source, AWS-style data lake, from raw object to partitioned table
S3-compatible MinIO, distributed PySpark processing, Parquet, partitioned PostgreSQL loading and Great Expectations checks — sized for about a million simulated transactions a day.
Simulated volume — not a measured production throughput
Archive
Older, smaller, still here.
Coursework, short internships and early work — for anyone who wants to trace the thread.
- 2024dzStance — StanceEval 2024Arabic stance detection, TF-IDF vs Sentence Transformers; ACL publication (co-author), 13th of 15 teams · Python · Sentence TransformersACL
- 2025SOAP solvency serviceSOAP/WSDL services, orchestration, TTL/LRU cache, Prometheus/Grafana — shared repository · Java · SOAP · WSDLGitHub
- 2024CERIST — video anomaliesVideo anomaly detection with CNNs and autoencoders (internship) · Python · TensorFlow · OpenCV
- 2024Crowd anomaly detectionOptical flow and classifiers on UCSD · OpenCV · scikit-learnGitHub
- 2024Arabic sarcasm detectionCorpus building and classification · Python · NLPGitHub
- 2024Card fraud detectionImbalanced classes, SMOTE, model comparison · scikit-learn · XGBoostGitHub
- 2024Genetic algorithm — TSPGenetic operators for the travelling salesman · PythonGitHub
- 2023MaktabatyDigital library, React front end and Django API · React · DjangoGitHub
- 2024Pharmacy web appPharmacy management and sales · React · Node.js · MongoDBGitHub
- 2023Oracle data warehouse & OLAPData warehouse and OLAP queries · Oracle · SQLGitHub
- 2025ZhengFen — SOLID & design patternsJava game structured with SOLID principles and design patterns · JavaGitHub