Mohamed CHAFIQ

– Designed and implemented a document-clustering system that automatically detects and groups document templates by visual and structural similarity. – Built a content-only analysis pipeline that combines computer vision and OCR to process heterogeneous, proprietary, and legacy documents at scale. – Applied unsupervised learning to identify unknown document types, accelerate template onboarding, and automate triage of unrecognized documents. – Enabled faster integration of new document templates into industrial OCR robots and computer vision pipelines. – Reconstructed and enriched document metadata (e.g., document type) to improve downstream OCR accuracy and document governance. – Industrialized the solution as an on-premise, reusable component, integrated into an existing production pipeline, and usable on demand. – Generated structured datasets with visual features and clustering indicators, supporting data quality teams and engineering workflows.

– Built an ML lead scoring system to rank prospects by conversion likelihood and improve sales prioritization. – Unified CRM + marketing + product signals into a clean modeling dataset and engineered intent/engagement features. – Trained and calibrated predictive models to output probability-based scores, with basic explainability (key drivers). – Integrated automated scoring into the sales workflow and added monitoring for drift/performance.

– Formulated objective functions to jointly optimize throughput, workload balance, and ergonomic risk across operators. – Incorporated ergonomic indicators (e.g., posture-based risk scores, task repetition, load exposure) as constraints or penalty terms in the optimization model. – Designed task-to-operator and task-to-station assignment strategies using Linear Programming and Integer Programming techniques. – Performed scenario and sensitivity analysis to evaluate trade-offs between productivity and ergonomic safety under varying production demands. – Enabled data-driven redesign of assembly line configurations to reduce operator strain while maintaining operational efficiency. – Delivered a reusable optimization component suitable for simulation or integration into decision-support systems for industrial engineering teams.

– Development of an AI model focused on Optical Character Recognition (OCR) engines to extract information from client documents. – Normalization of OCR results to tailor them to the team's needs. – Development of a solution to address the issue of line detection in a document. – Dockerize and deploy the solution.

– Exploration of recent models and advanced algorithms for video classification. – Implementation of the most widely used recent methods in the field of automated human activity recognition. – Conducting a comprehensive comparative study of the performance of different video classification models.