Randrianarisoa

Built the first version of an AI-powered tourism analytics platform (Touria), translating tourism-flow business needs into backend services, data pipelines, analytical dashboards, secure REST APIs, and actionable AI-generated insights. Developed Django backend services and REST APIs with centralized business logic and role-based access control (RBAC). Designed and maintained Apache Airflow ETL pipelines for reliable data ingestion, transformation, and delivery to analytics and predictive models. Integrated LLMs with structured prompts to produce consistent French explanations for visit trends, overnight stays, geographic distribution, demographics, and predictive factors. Managed Linux infrastructure on Google Cloud and automated Docker deployments via Jenkins CI/CD to support high availability.

Developed a production data platform using Java Spring Boot and FastAPI, integrating sales, orders, and inventory data from Facebook and Instagram to drive merchant decision-making and operational workflows. Built backend services and integrations to consolidate fragmented commercial data into a unified platform. Implemented Google ADK-based AI agents for natural-language business data analysis, intelligent order extraction from customer conversations, and commercial process automation. Added Prophet-based inventory forecasting to reduce stockout risk and improve replenishment planning. Enhanced PostgreSQL data architecture and used Redis for high-concurrency streaming analytics to meet sub-second response times for social media data.

Contributed to a “Database-first” scaffolding tool that reduced MVP development time from days to minutes by auto-generating full-stack codebases. Scaled and supported the open-source engine for MySQL-based C# and Java generation, resolving critical database connection issues in the IntelliJ plugin. Authored complex ASP.NET search logic and API handlers to improve robustness and production-readiness of generated code.