Ayoub ETOULLALI

Freelance Data Scientist, Analyst, AI Engineer, Automation & Scraping

Worldwide • Freelance

Dec 2022 - Present • 3 yrs 10 mos

Delivered 50+ commercial technical solutions spanning Generative AI, predictive modeling, high-throughput web scraping, and enterprise automation:  Applied AI & Agentic RAG Systems: Architected document retrieval assistants using LangChain and LlamaIndex. Built scalable AI content engines (GPT-4o, Claude) via n8n and Flask to process databases of 30M+ records into targeted media. Deployed Vision-AI clinical workflows, multilingual chatbots, ATS resume parsers, and RLHF evaluation pipelines.  Machine Learning & Advanced Analytics: Engineered explainable clinical risk models (XGBoost, Random Forest) with SHAP interpretability and Streamlit dashboards. Developed e-commerce time-series demand forecasting models paired with Power BI, association-rule recommender engines (Apriori), sports valuation algorithms, and telemetry anomaly detection systems.  Web Scraping & Data Engineering: Architected distributed data ingestion frameworks processing over 9M public records and automated tender alerts (AlertesMarches). Built resilient scraping pipelines with dynamic proxy rotation, anti-detect browser automation (Playwright, Selenium, Incogniton), and robust JSON/relational ETL validation systems.  Workflow Automation & Integration: Deployed autonomous publishing loops, B2B lead generation engines, and operational workflows using Make.com, self-hosted n8n, and custom Python scripts. Integrated Shopify inventory webhooks, social media APIs, and automated file-watchers. • Core Stack: Python (Pandas, Scikit-Learn, XGBoost), LangChain, LlamaIndex, OpenAI/Claude APIs, Streamlit, Power BI, Playwright, Scrapy, n8n, Make.com, Flask, PostgreSQL, MySQL.

Integration / Middleware Engineer

STRATAGEMS • Freelance

Jun 2026 - Jul 2026 • 1 mo

 Architected and Maintained Middleware: Designed, built, and maintained real-time integration flows connecting eCommerce platforms, ERPs, and Product Information Management (PIM) systems to seamlessly synchronize orders, pricing, inventory, and customer data.  Engineered Resilient Data Pipelines: Developed defensible, fault-tolerant integrations utilizing unique keys and idempotency to ensure clean retries, data deduplication, and zero data loss during upstream system failures or out-of-order events.  Implemented Strict Security Protocols: Enforced end-to-end security across all integration layers, managing OAuth credentials, encrypting data in transit, applying least-privilege access, and ensuring sensitive data (PII) was never leaked into system logs.  Managed Operational Reliability: Anticipated edge cases and operational needs by building comprehensive monitoring, alerting, and replay/recovery mechanisms, alongside writing detailed runbooks for rapid incident resolution.  Executed API & Legacy Integrations: Connected diverse business systems utilizing a wide range of protocols and interfaces, including REST and GraphQL APIs, webhooks, and SFTP feeds.  Ensured Production Quality & Debugging: Conducted rigorous testing against complex edge cases (e.g., empty payloads, partial failures) and actively debugged production issues using SSH, SQL, and log analysis to quickly identify and explain root causes. Tools & Technologies: - Programming & Scripting Languages: TypeScript, JavaScript, SQL - APIs, Protocols & Authentication: REST APIs, GraphQL, OAuth, Webhooks, SFTP - Middleware & iPaaS: Patchworks, Custom-coded integration layers - Systems & Platforms (Integration Endpoints): eCommerce (Shopify), Enterprise Systems (ERP, PIM, OMS) - DevOps, Operations & Debugging: SSH, Log Management & Analysis, Proactive Monitoring & Alerting systems - AI & Productivity: Claude (AI-assisted development)

Data Scientist, AI Engineer & Automation Specialist

Xziry • Temps plein

Apr 2025 - Apr 2026 • 1 yr

 Agentic RAG & MCP Architecture: Designed and deployed an advanced Agentic RAG architecture orchestrated via FastAPI, Redis, and PostgreSQL, featuring LLM function/tool calling and security-hardened prompts. Extended the platform using Model Context Protocol (MCP) to standardize context handling while preserving peak performance standards.  Hybrid Conversational AI: Implemented a production-grade hybrid chatbot using Rasa, combining NLU for structured FAQ handling with LLM-powered natural language generation to manage complex queries and end-to-end booking workflows.  Evaluation & Observability: Built a multi-metric evaluation pipeline covering RAG quality, tool-calling accuracy, LLM output assessment, and security auditing, visualized through Streamlit. Integrated Grafana dashboards for real-time system performance tracking.  End-to-End QA Automation: Engineered a Playwright-based QA framework to automate core user journeys (authentication, performance benchmarks, booking workflows), generating structured execution reports and screenshot-backed audit trails.  Solution Architecture & Documentation: Partnered with Product Owners on technical solution designs and feasibility assessments. Authored layered technical specifications and stakeholder summaries to bridge business and engineering needs. Tools & Technologies - AI & LLM: Rasa, OpenAI API, Agentic RAG, Model Context Protocol (MCP), LLM Tool Calling, Prompt Engineering - Backend & Data: Python, FastAPI, Redis, PostgreSQL - Testing & Observability: Playwright, Pytest, Streamlit, Grafana