Omar Nouiri

• • Scoped and deployed end-to-end AI use cases in insurance: prediction, scoring, document intelligence — from conception to production on AWS

• • Architected and optimised a production RAG system (hybrid search, reranking, prompt engineering) — cutting response latency by 10s and improving accuracy by 20%

• • Built the first agentic flow POC on top of the RAG system using tool use / function calling, enabling the LLM to dynamically invoke external tools

• • Designed and delivered a PII redaction pipeline for insurance and financial documents (names, emails, phone numbers, free text) combining LLMs, rule-based extraction, and AWS serverless architecture (Textract, Step Functions, Lambda) — enabling secure document sharing at scale

• • Currently developing a PII extraction system to automatically identify and classify sensitive entities across heterogeneous unstructured document types

• • Active R&D on frontier LLM techniques and agentic architectures; presented new AI use cases to business stakeholders

• • Won the national Fintech Catalyst challenge by building a Darija-dialect conversational banking assistant (RASA + custom NLP models) — demonstrated to clients at GITEX

• • Deployed French and Arabic NLP transcription models achieving 95% accuracy, integrated into a banking app serving 10,000+ active users

• • Built Airflow ETL pipelines processing 500,000+ job postings; applied BERT, GPT, LLaMA for classification, skill extraction, and semantic recommendations

• • Trained end-clients on deployed AI solutions; built automated Power BI dashboards for executive reporting

• • Built the first data scraping and analysis pipelines covering 100,000+ Moroccan job listings

• • Designed and optimised a MySQL schema for structured storage and large-scale insight extraction