António – Python, AWS, ETL

An AI-powered B2B procurement platform that automates supplier and product matching, purchasing workflows, and business interactions through WhatsApp and SMS. The solution combines conversational AI agents, multi-agent orchestration, and graph-based retrieval to support complex procurement processes, with a web application for managing procurement workflows.

• Architected and delivered a production AI-powered procurement platform, enabling automated B2B interactions through WhatsApp and SMS;
• Designed and implemented multi-agent orchestration using Semantic Kernel, GPT-4, RAG, and tool calling to automate complex procurement workflows;
• Designed and optimized Neo4j knowledge graph models and Cypher queries to support intelligent supplier and product matching;
• Implemented GraphRAG with Neo4j vector and full-text indexes, combining semantic and lexical search for improved entity resolution;
• Developed and deployed a Vue 3 single-page application and Web APIs, integrating an AI chatbot into procurement workflows;
• Built and maintained cloud infrastructure and CI/CD pipelines using AWS, Kubernetes, and Terraform, incorporating disaster recovery controls.

An AI-powered regulatory intelligence platform supporting device manufacturing certification by processing regulatory standards, requirements, and testing strategies. The solution uses LLM-based data annotation and validation to support regulatory information processing, with a focus on data quality, consistency, and reliability.

• Developed data engineering components for an AI-powered regulatory intelligence platform supporting device manufacturing certification;
• Designed and implemented an automated quality monitoring system for LLM-generated annotations using Python and hexagonal architecture;
• Built scoring and validation pipelines to evaluate AI-generated annotations against reviewed reports and identify quality discrepancies;
• Defined and implemented a new data architecture covering data ingestion, transformation, governance, and modelling to support regulatory intelligence workflows.

A data analytics and decision intelligence platform designed to help businesses make data-driven decisions and improve performance. The engineering work focused on modernizing legacy data infrastructure, establishing scalable data processing workflows, and improving the reliability and maintainability of the company's analytics ecosystem.

• Led Data Engineering, ML Engineering, and DevOps functions, restructuring engineering teams into multidisciplinary squads to improve collaboration and delivery;
• Proposed and drove the adoption of dbt, replacing legacy SQL scripts running on Talend with modern data transformation workflows using Snowflake and dbt;
• Directed the migration of legacy data infrastructure to a medallion architecture built on Snowflake, Apache Airflow, and dbt;
• Established engineering delivery standards and mentored engineers and team leads on code quality, technical practices, and execution.

A rail travel retail platform that enables passengers to purchase affordable ticket upgrades through a mobile application. The platform's data infrastructure supports real-time analytics, pricing and inventory forecasting, and reporting for business partners. The engineering work focused on building scalable cloud data pipelines, modernizing infrastructure, and enabling data-driven decision-making.

• Designed and implemented a cloud-based data lake and warehouse architecture on AWS using Amazon S3, Athena, and Redshift;
• Developed and orchestrated large-scale Python ELT workloads using Apache Airflow and Kubernetes Operator;
• Built a real-time analytics platform using Apache NiFi, Amazon SQS/SNS, and Snowflake, delivering the solution within three months;
• Introduced an event-driven architecture with a centralized event taxonomy to standardize data collection across frontend and backend systems;
• Led Kubernetes adoption using Amazon EKS and Infrastructure as Code, deploying a highly available Airflow environment with Helm, disaster recovery capabilities, and CI/CD pipelines;
• Implemented data anonymization mechanisms for ELT pipelines processing personally identifiable information (PII) and indirectly identifying data;
• Developed ML training pipelines for pricing and inventory forecasting using Apache Airflow and AWS Batch, with DynamoDB supporting model output delivery;
• Designed and implemented automated reporting services, reducing partner integration effort by approximately 90%.