Zakariae El Mernissi

👨‍🔬 Member of the GenAI Lab, Building High-Impact Generative AI Solutions 🧠 AI-Powered HR Platform Architected a production multi-agent GenAI copilot using Google ADK, automating resume parsing and intelligent candidate-role matching, reducing manual screening time by ~50%. - Translated HR requirements into scalable AI specifications, defining agent roles, data flows, and conversation orchestration. - Engineered prompt engineering strategies (zero-shot, few-shot, chain-of-thought) for context-aware, regulation-compliant outputs. 🌍 Supply Chain Analyzer: Aerospace Industry - Built a cloud-native conversational AI system on AWS Bedrock for real-time supply chain constraint analysis,~60–65% faster query resolution vs. manual review. - Architected a full RAG pipeline with Amazon Titan embeddings and OpenSearch vector store. - Implemented Claude via Bedrock as the reasoning layer with cited document references. - Designed secure serverless architecture (Lambda, API Gateway, S3, CloudFront, Cognito). 💊 CSV Platform: Pharmaceutical Industry - Engineered a multi-agent GenAI system using CrewAI for document analysis and compliance case management reducing validation cycle time by ~45–50%. - Translated GxP regulatory requirements into structured technical specs, defining agent responsibilities and validation logic. - Established CI/CD via GitHub Actions for automated testing and delivery to Azure Web Apps.

Concevoir et développer un système informatique qui intègre des technologies d'intelligence artificielle générative implique : - Optimiser l'intégration des nouveaux collaborateurs et automatiser l'accès aux applications. - Générer la rétro-documentation et les spécifications fonctionnelles à partir du code existant. - Produire les documents de transfert de connaissances à partir des vidéos. - Améliorer la recherche d'informations et la génération de résumés de CV.

Built comprehensive video-to-audio processing pipeline supporting YouTube URLs and direct video file formats. Implemented advanced audio preprocessing with noise reduction using Deep Filternet for enhanced accent detection accuracy. Integrated Wav2Vec2 transformer model via Hugging Face for sophisticated accent classification (British, American, Australian, etc.) Created confidence scoring system with AI-powered natural language explanations using OpenRouter API integration. Designed user-friendly interface with real-time processing capabilities and detailed analytics dashboard. Automated multimedia content analysis with seamless video download and audio extraction workflows

Developed an intelligent AI agent system for automated commercial property ownership research and due diligence Built multi-source data integration pipeline connecting public records, property databases, and legal document repositories Implemented entity extraction and relationship mapping using GPT-4 for property ownership chains and corporate structures Created automated web scraping workflows with intelligent data validation and cross-referencing capabilities Designed knowledge graph representation using Neo4j to visualize complex ownership relationships and property portfolios Built semantic search capabilities for property records using vector embeddings and contextual retrieval Developed comprehensive reporting system with automated property analysis, risk assessment, and ownership timeline generation Integrated compliance checks for regulatory requirements and due diligence standards