Currently an Extern @PfizerI build production-ready ML pipelines, statistical models, and portfolio optimization tools using Python and PyTorch, with hands-on experience in computer vision and NLP.
Portfolio highlights include a portfolio-optimization trading system that won a hackathon, a low-latency CV inference engine, a tweet entity linking pipeline, and a Pfizer externship prototyping
.jpg)
Pfizer · ✅ Verified by Extern · ⏱️ In progress
Prototype AI-powered document intelligence with Pfizer—using OCR, LLMs, and RAG to automate real enterprise PDF workflows and build a standout portfolio project.
Won first place (Compact Hackathon) among all competing teams by delivering an end-to-end quantitative application for portfolio optimization and algorithmic stock trading recommendations, built and demoed live within the hackathon time limit
I build production-ready ML pipelines, statistical models, and portfolio optimization tools using Python and PyTorch, with hands-on experience in computer vision and NLP.
Applied Mathematics & Statistics student with internship experience spanning banking data science and production-grade machine learning. Proficient in statistical modeling, building production-grade ML pipelines, and portfolio optimization using Python and PyTorch. Skilled at communicating quantitative findings cleanly to cross-functional stakeholders.
Externships
Pfizer Advanced: AI-Powered Document Insights & Data Extraction Externship
In progress
Experience
Final Year Research Project (PFE) 6 Incoming
Centre de Recherche en Math´ematiques Appliqu´ees (CRMA) · Starts Sep. 2026
Machine Learning Engineer Intern
Numedia Mind Technologies (NMT) · Feb. 2026 6 Jun. 2026
Data Science Intern
Banque Ext e9rieure dAlg e9rie (BEA) · Jul. 2025 6 Aug. 2025
Education
National Higher School of Mathematics (NHSM) Algiers, Algeria
Master of Engineering in Statistics and Applied Mathematics · Class of 2027
Skills
✅ Verified by Extern · ⏱️ In progress
Prototype AI-powered document intelligence with Pfizer—using OCR, LLMs, and RAG to automate real enterprise PDF workflows and build a standout portfolio project.
The externship explored AI-driven document intelligence for enterprise PDFs using OCR, large language models, and retrieval-augmented generation. The work involved researching core LLM concepts and demonstrating how those components integrate into document-extraction pipelines. Deliverables documented model behavior, architectural considerations, and example integration patterns.
I researched foundational LLM concepts and explained how they relate to document-intelligence pipelines, producing written notes and summarized diagrams used across the project.
Won first place (Compact Hackathon) among all competing teams by delivering an end-to-end quantitative application for portfolio optimization and algorithmic stock trading recommendations, built and demoed live within the hackathon time limit
Built a production-grade multimodal NLP pipeline to disambiguate and link tweet entities to Wikipedia counterparts, combining DNNs, XGBoost, and SVMs in a stacked ensemble with model evaluation across all components Achieved accuracy on par with commercial APIs (AIDA, TagMe) while delivering inference in under 100 ms, outperforming both on latency
Engineered and deployed a real-time computer vision system for vehicle speed detection to a confidential Algerian company; the production system achieved sub-millisecond inference latency within a multithreaded architecture Exposed model predictions through a RESTful FastAPI service with strict Pydantic validation, containerized with Docker for scalable cloud deployment; optimized the YOLOv8 serving pipeline to sustain concurrent video streams without frame loss Note: linked GitHub repository is an open exploratory prototype used to test and get familiar with the underlying tech stack 6 it does not reflect the final proprietary production system
Implemented a functional relational database engine from scratch in C, replicating core SQLite features including B-Tree indexing, a SQL parser, and full CRUD operations Enforced memory-safe access patterns throughout; achieved benchmark query performance within comparable range of standard SQLite
