Saad Abidi

Ouicare Mission — GenAI (04/2025 – Present) • Led technical delivery of a unified, multi-agent AI-native ERP system, coordinating two Data Scientists and a Product Manager while owning its architecture and key components. • Started with a POC enabling natural-language Snowflake queries through AWS Bedrock and Cortex Analyst. • Evolved the POC into a production system across 300+ tables using LangGraph, AWS AgentCore, pgvector retrieval, AWS CDK CI/CD, and guardrails; supports data queries, ERP API actions, and step-by-step guidance for 100k+ users. Internal Data Platform (Codex) (02/2023 – Present) • Contributed to the development and adoption of an internal Data & AI platform; delivered AI/ML POCs and LLM systems across conversational analytics, OCR, predictive modelling, and real-time dashboards. • Built an end-to-end event-intelligence POC for the City of Marseille: collected public Instagram and X content, classified local news and events with Qwen 3.5, and delivered a Streamlit application. • Designed and integrated an end-to-end DeerFlow agent harness using skills, tools, and MCP servers to operate a cloud-agnostic, Kubernetes-native data platform: tables, data queries, pod/log inspection, code review, commits, and GitLab merge requests across ClickHouse, PostgreSQL, Prometheus, Airflow, Superset, Grafana, and GitLab, with human approval for sensitive actions. • Led client enablement through integration, optimisation, and architecture recommendations. Accor Hotels Mission — Operations Optimization (10/2023 – 04/2025) • Defined KPIs and thresholds with business teams for outbound information flows to external partners; built a predictive anomaly-detection system that reduced critical incidents by 87%. • Designed dynamic data-flow mapping to improve system observability and troubleshooting. Leadership & knowledge sharing: Mentored junior Data Scientists and interns; delivered Data & AI training sessions and organised internal hackathons.

• Developed and deployed a T5-based NLP system that generated search queries from product descriptions, improving product-to-query matching and search relevance. • Led the project end-to-end, including literature review, data preprocessing, model optimisation, and BLEU/ROUGE benchmarking.

Developing a web application to manage the company's purchasing workflow.