Ayoub El Houdri

I designed and deployed a complete LinkedIn prospecting automation solution, combining artificial intelligence and advanced data processing.

This solution is based on a multi-agent system capable of automatically generating personalized prospecting messages, perfectly tailored to targeted profiles.

To maximize campaign relevance, I segmented LinkedIn profiles based on their field of interest and engagement level, by analyzing their publications and interactions. This fine segmentation improves targeting quality and increases conversion rates.

The solution also integrates a vector search engine to semantically identify qualified leads, ensuring intelligent and relevant prospecting.

Finally, I implemented LLMOps pipelines to ensure the robustness, continuous monitoring, and scalability of models in production, for optimal performance and simplified maintenance.

Website: https://www.waalaxy.com/

I designed and developed a multimodal AI agent based on a RAG architecture, intended to automate the generation of CIR (Research Tax Credit) reports, and also applicable to CII (Innovation Tax Credit) schemes.

This solution, designed for large consulting and auditing firms, allows for semantic and intelligent analysis of various data sources—R&D technical documentation, scientific bibliographies, patents, and internal financial data (budgets, timesheets)—to produce a structured, argued report compliant with regulatory requirements.

I worked closely with R&D, Finance, and Compliance teams to structure documentary production workflows and industrialize report generation. Through this approach, the preparation time for each file was reduced from 40 hours to less than 4 hours, while ensuring complete traceability, enhanced documentary consistency, and high compliance, with a very low hallucination rate.

The solution relies on a modern infrastructure including a monitoring system (LLMOps), ensuring the supervision, scalability, and maintainability of the system in production, within an intuitive, secure platform compliant with European data residency and protection standards.

Website: https://www.finexov.com/

Development and production deployment of scraping pipelines to extract, structure, and enrich data from over 15 websites in the furniture and retail sector. The collected information concerned resellers: contact details, website, social networks, contacts, owner information, etc.

The scripts were developed in Python, with the implementation of anti-scraping protection bypass mechanisms (IP rotation, user-agents, delay management, proxies). The data was then cleaned and centralized in a data warehouse (PostgreSQL), designed for analysis, quality control, and export.

This data was used to feed a market study and a competitor analysis, identifying the geographical distribution, commercial positioning, and network coverage of industry players.

Website: https://fr.cornilleau.com/