Guillaume Rakotonjanahary Tsantaniaina Amboarasoa

🟡 Data Warehouse Design, ML & Visualization: Built a biometric KPI data warehouse by defining fact and dimension tables from OLTP PostgreSQL schemas and modeling them in a BigQuery OLAP star schema. Created materialized views for fast Looker dashboards and used BigQuery ML (ARIMA+, ML.FORECAST) to forecast kit usage and trends . 🟡 No-code e-commerce product recommendation app with NLP and image recognition on GCP (Vertex AI, BigQuery) using Python FastAPI/React TypeScript. RAG-based chatbot with configurable modes (image/text/combined), admin dashboard with analytics and customization, Google Sheets integration for catalog management. Used by pre-sales/marketing teams, <3s response time. 🟡 SQL generation model benchmarking: Evaluated SQLCoder-7B-2, GPT-3.5/4, Claude for architecture, hallucination mitigation, BigQuery syntax, multi-cloud deployment (HuggingFace/AWS/GCP), Spider benchmark validation. Delivered production-ready technical framework. 🟡 Innovation Cell: Benchmarked AWS cloud services and GCP equivalents including SageMaker, Bedrock, Forecast, Lex, Polly, Rekognition, Textract, Glue, S3, RDS, Redshift, Lambda. Created comprehensive comparison table with service descriptions, technical properties, strengths, limitations, pricing models, and use cases. 🟡 Real-time Biometric Enrollment KPI Platform: Analytics solution with PostgreSQL data replicated via Airbyte CDC to BigQuery, transformation with DBT SQL , visualized in Looker dashboards for instant monitoring of mobile biometric kit performance in identity verification. 🟡 BI Agent Chatbot: RAG-based text-to-SQL solution using Python, LangGraph, LLMs, FastAPI, React TypeScript. Retrieves SQL patterns, DDL, metadata for multi-query execution and instant biometric enrollment insights. 🟡 Support Agent Helpdesk: LangGraph/RAG-powered system for kit usage guidance, troubleshooting, and technical support using process documentation for context-aware biometric enrollment assistance.

🔵 Consistently create comprehensive project backlogs that translate client requirements into functional and technical specifications, including: ● Use Cases and User Stories with detailed functional requirements ● Business Rules (Règles de gestion) defining operational constraints and logic ● Technical Cases outlining implementation specifications ● Development Effort Estimation with detailed time allocation (in days) for each user story ● Cloud Cost Estimation for infrastructure and deployment resources ● UML Diagram Creation including: ( Deployment Diagrams , Activity Diagrams , Sequence Diagrams , Use Case Diagrams , Class Diagrams , Cloud Component Architecture Diagrams Etc ... ) 🟣 Designed, developed, and deployed a multilingual conversational AI assistant as both web and WhatsApp applications, integrated with Python FastAPI. Utilized advanced NLP techniques to efficiently handle information requests (flight status, navigation, services) and serving also as a helpdesk tool for airport staff. Implemented operational support using Python (FastAPI), React (TypeScript), Google Kubernetes Engine (GKE) for scalable deployment, and BigQuery for user query analysis and data structuring. 🟣 Participated in designing a generalized and reusable architecture for chatbot creation using Draw.io with UML methodology, implementing good design patterns with LangGraph, API access, and external service integration including knowledge base with RAG (Retrieval-Augmented Generation). 🟣 Developed an intelligent product catalog system for image-based product search in e-commerce websites using Python and generative AI. 🟣 Implemented a distributed Big Data clustering project for online e-commerce image data using Apache Spark and Google DataProc.

🔴 Developed a hybrid RAG (Retrieval-Augmented Generation) architecture for chatbot systems, combining vector search and knowledge graph technologies. Implemented comprehensive search capabilities for information extracted from documents and images stored in cloud infrastructure as a Master's Thesis Project in Artificial Intelligence. 🔴 Developed an ETL pipeline for vectorizing 105,000 images into BigQuery using Google DataFlow and Apache Beam. Built clustering and recommendation systems for search optimization, enabling vector similarity searches across both images and text descriptions. 🔴 Participated in developing an automated Business Intelligence assistant 🔴 Participated in developing an automated resume/CV recommendation assistant using machine learning models, built with Python FastAPI and ReactJS to match candidates with relevant opportunities. 🔴 Consistently create comprehensive project backlogs that translate client requirements into functional and technical specifications, including: - Use Cases and User Stories with detailed functional requirements - Business Rules (Règles de gestion) defining operational constraints and logic - Technical Cases outlining implementation specifications - Development Effort Estimation with detailed time allocation (in days) for each user story - Cloud Cost Estimation for infrastructure and deployment resources - UML Diagram Creation including 🔴 Participated in developing an automated product counting tool for supermarket inventory management using Computer Vision technologies, implemented with Python TensorFlow, YOLO V8 library, and OpenCV for real-time object detection and counting. 🔴 Developed an automated AI assistant for natural language product search in a flowers e-commerce platform and a water assistance and purification services website, utilizing Python FastAPI , React Typescript , Bigquery Vector Search RAG and generative AI to enable intuitive search queries and product recommendations.

- Designed comprehensive technical architecture and specifications for new data collection modules supporting malaria bulletin generation with dashboard solutions for malaria data visualization to enhance availability and interoperability of quality health information systems between the Ministry of Health and Program Office Malaria information systems (Microsoft Access and DHIS2). Solution leveraged React.js, Chart.js, and LLM models for natural language to SQL query conversion with DHIS2 API management (CRUD operations) and external custom data source integration. Collaborated with multidisciplinary team including Data Scientists, SME Malaria Advisors, Program Officers, ICT Health Informaticians, and Health Informatics Advisors. - Architected and deployed DHIS2 instances in development and pre-production environments with comprehensive module configuration for data entry workflows. Developed custom forms and created data elements and key performance indicators (KPIs) including WASH indicators, PSERAN 2022-2026 indicators, and PARN/HORS PARN indicators collected at UPNNC level. Implemented using Java DHIS2 framework, Tomcat server, PostgreSQL database, and Linux server configuration. -

- Participated in the organization and creation of review documents and slides showcasing the activities carried out for the closure of the PMI Measure Malaria program. - Assisted in organizing the farewell event marking the end of the PMI Measure Malaria program, attended by partners and entities including representatives from the Ministry of Health and other collaborating organizations. - Assisted in planning a smaller farewell event for the completion of the Resident Advisor’s assignment in Madagascar, marking the end of their local activity. - Engaged in independent research to strengthen knowledge in Health Information Systems and related development areas. - Developed personal projects to enhance creativity, skills, and professional growth. - Conducted a feasibility analysis for the creation of a geolocation app with advanced features through remote discussion with a key contact who had already consulted other partners. - Undertook personal learning and training in new technologies, including AI and advanced data science, and began the Master 2 program to further academic and professional development. - Balanced professional development with personal renewal during the project gap, ensuring preparedness for future opportunities.

- Provide training and technical support to health sector civil servants as a specialized trainer in the use of data collection tools, including KoBoToolbox for form design, structuring, and visualization, as well as the development of associated technical documentation. Serve as assistant trainer and pedagogical facilitator during QGIS cartography training sessions, providing personalized technical support to resolve difficulties encountered by participants during the sessions. - Provided technical support for expanding scorecard and dashboard mobile application visualization tools across selected regions in Madagascar. - Contributed to vaccination site visualization mobile application version development using React Native framework. - Data Visualization and Health Facilities Geolocation Application - This USAID PMM Program tool streamlines vaccine stock management and healthcare facility navigation by providing comprehensive vaccine inventory visualization across healthcare facilities with search capabilities by vaccine type and date. The application integrates geolocation features to display healthcare and vaccination centers on interactive maps, offering current location tracking and turn-by-turn directions while supporting both private and public health facilities - Development of the Digital Health Library - This USAID PMM Program project creates a comprehensive digital repository for the Ministry of Health's SARGEC Department, focusing on digitizing and organizing health-related documents, research papers, and informational resources specifically tailored to departmental needs. The initiative enhances research document management through improved accessibility, categorization, and retrieval systems while integrating with existing Health Information Systems to streamline information flow and support evidence-based decision-making for health policies and practices within the Health Information Systems Directorate.

- Automatic Detection of Covid 19 Data Anomalies - Developed from September 2022 for USAID's PMM Program, this project enhances COVID-19 data integrity through automated detection of common data issues including duplicates, missing values, outliers, and non-applicable entries. Implemented as a DHIS2 platform module using React.js and Node.js, the system automatically identifies repeated data entries, detects gaps in data collection, and manages non-applicable data to maintain dataset consistency and reliability for accurate public health assessments - Contributed to the design and development of quality control modules for DHIS2 data validation and parameter optimization. - Developed comprehensive mailing list management application with individual and group email functionality, specifically designed for Gmail integration using Gmail API and Google Cloud Service Account. Built using Electron.js, React.js, and SQLite database. - Provided technical assistance to project teams for activity implementation at peripheral levels according to established project planning schedules. - Supported project teams in managing and maintaining COVID-19 vaccination data systems. - Contributed to supervision activities and data regularization processes to ensure data integrity and compliance. - Facilitated user and stakeholder engagement workshops, participated in meetings and workshops at both central and peripheral levels with PMM project support. - Participated in data entry processes, activity report capitalization, and maintained attendance records during project activities.