About
The last mile is the work.
I am a fourth year Data Science Engineering student at Abdelhamid Ibn Badis University. I build AI systems end to end: data, modeling, retrieval, evaluation, deployment, and the interface. I care more about baselines and failure modes than leaderboards.
Approach
Start with a baseline. If the method cannot beat a simple one, add complexity only when the metric shows it. Keep preprocessing inside the fold. Split by time when order matters.
Treat uncertainty as a product decision. A review flag is better than a false confident answer. In voice, latency is the product. In retrieval, the retriever is measurable before the generator.
Write down limitations. If a number is missing, it stays as a pending state until it exists. The site enforces that.
Now
RAG retrieval evaluation
In progressFreezing the eval set and comparing BM25 with the dense retriever.
ECG review flag calibration
In progressTurning signal quality into a threshold a reviewer can trust.
This portfolio
ShippedFilling missing facts and running the deployed audit.
Education
Data Science Engineering
Abdelhamid Ibn Badis University
Fourth year of five / Remote
Internships and certifications are pending. See the resume for the current state.
GitHub
Set GITHUB_USERNAME in the environment to show live GitHub stats. No data is fetched in this build.