At Algeria Partners - Temacina, I designed a multi-step pipeline to extract and structure contact data from more than 300 images. I built a hybrid OCR system using PaddleOCR, EasyOCR, and Tesseract with OpenCV preprocessing, then added normalization and deduplication to produce structured JSON.
For my Advanced RAG System for Intelligent Document Retrieval project, I engineered a retrieval pipeline with ChromaDB and cross-encoder reranking, and developed query decomposition with LangChain. I also built an evaluation framework with Ragas and deployed the modular pipeline as a FastAPI microservice.
I implemented a GPT-2 inference engine from scratch in Python and NumPy, including tokenization, attention, KV caching, and text generation. I’ve also built an aircraft hydraulic-system simulator and machine-learning pipeline for predictive risk analysis, while creating technology and AI content and organizing student tech events with GDG.
