Wladyson

Additional Extras

Financial Platform

• The project consists of building several microservices responsible for account management, financial transactions, payments, treasury, accounting ledger, auditing, reconciliation, risk, scoring, digital channels, notifications, identity and security (IAM/KYC), orchestration of financial processes, integration with external systems, batch processing, and operational back office, among others. The solution operates on a Java architecture with asynchronous processing and communication via Kafka, ensuring consistency with the SAGA standard and immutable event logging for regulatory purposes. The infrastructure is built with Terraform, using EKS, RDS, VPC, Vault, CI/CD with GitHub Actions and GitOps with ArgoCD, as well as observability with Prometheus, Grafana, and Alertmanager. In the initial phase, the environment's foundations and the accounts module are implemented, including complete CRUD and transactional integration with PostgreSQL. Following this, the transactions module is built, responsible for PIX, TED, and DOC transfers, using events in Kafka and guaranteeing exactly-once processing. Next, the card and reconciliation modules are developed, with daily processing routines and real-time workflows. The final phase includes the immutable audit system, performance optimizations, and resilience testing with controlled failures in Kubernetes, along with the completion of the SLO and SLI dashboards for monitoring the behavior of the microservices. The accounting closing processes perform nightly calculations and aggregations, while the audit trail permanently saves all events to comply with regulatory requirements.
  

Asynchronous Payment Gateway System

• This project is designed for high-concurrency, low-latency scenarios, inspired by systems like PIX. The platform receives payment requests and executes processing in a completely asynchronous manner, ensuring logical atomicity and eventual consistency between accounts. The architecture is event-driven and composed of two main microservices: a Gateway Handler, responsible for ingesting, validating, and publishing payment events, and a Transaction Processor, responsible for executing debits and credits. Coordination between services is performed using the SAGA pattern via choreography, allowing for automatic compensation in case of partial failures. Communication between components occurs via Apache Kafka, ensuring reliable delivery, ordering, and fault tolerance during processing. The system runs on Kubernetes, guaranteeing orchestration, scalability, and high availability. The Gateway Handler was developed in Java with Reactive Quarkus, using Mutiny for non-blocking programming, allowing thousands of simultaneous connections with low latency. Critical performance metrics, including P99 latency, are continuously monitored using Prometheus. The persistence layer uses PostgreSQL as the transactional database, provisioned via Terraform, with explicit configuration of isolation levels and lock control to preserve transactional integrity in concurrent scenarios.
  

Account Reconciliation Engine (Batches and Streams)

• An application designed to compare large volumes of internal transactions with bank statements, combining batch and real-time processing to ensure balance accuracy and reliable identification of discrepancies. The system performs batch reconciliation through complex jobs implemented in Java with Spring Batch, optimized for reading, processing, and writing large files, including bank statements up to 10 GB. The recurring execution of these processes is orchestrated via Kubernetes CronJobs, ensuring predictability and load isolation. For continuous reconciliations, the engine uses Kafka Streams to process financial events in real time, allowing for immediate reconciliation of transactions, reversals, and corrections as events are published, eliminating the exclusive dependency on daily batch processing. Raw files are stored in AWS S3, ensuring durability and traceability, while consolidated reconciliation results are persisted in AWS Redshift, enabling historical analysis and large-scale analytical queries. The entire infrastructure was provisioned as code with Terraform, ensuring reproducible and consistent environments. The system's observability was structured with Datadog, monitoring worker health, processing throughput, and Kafka queue lag, allowing for proactive detection of bottlenecks and operational failures.
  

Suprema Admin - Healthcare Administrative Automation System

• This data automation and processing system is designed for the administrative management of clinics and healthcare institutions. It handles large volumes of unstructured information and integrates it with official systems automatically, accurately, and in an auditable manner. The system ingests and interprets data from completely unstructured sources, including PDFs, spreadsheets, patient records, medical reports, manual forms, and various other reports. From these inputs, the platform extracts, organizes, and structures the information, generating standardized spreadsheets and detailed reports with appointment times, clinical observations, and data ready for analysis and validation, eliminating the need for manual data entry. In addition to data organization, the platform performs complete automation of data entry into official systems, recognizing fields, validating information, and filling out external forms with high precision. In scenarios of data inconsistency or absence, the system issues alerts, blocks the submission of incorrect information, and automatically highlights problematic fields, ensuring compliance and reducing operational errors. The solution also automates the identification of spreadsheets and reports required by different regulatory bodies, extracting relevant information and submitting it according to specific deadlines and formats, eliminating manual processes and email submissions. This workflow ensured standardization, traceability, and continuous regulatory compliance. As an operational result, the system automatically processed 12,526 data entry procedures in just 5 days, a volume that previously required between 45 and 50 working days of manual work by a team of approximately 6 people. Automation drastically reduced processing time and freed up the administrative team for strategic activities.
  

Limits and Fraud Monitoring API

• This project focuses on real-time validation and decision-making to prevent fraud and limit overruns during transactions. The goal was to create a high-throughput Java API that, with each received transaction, queries the customer's history and applies complex business rules—such as daily limits and usage frequency—in milliseconds. To ensure fast decisions, Redis (via Caffeine as a local cache + remote Redis) was used to store limits and recent customer data in memory, maintaining response times below 50ms. The Redis cluster was provisioned with Terraform using AWS ElastiCache. Business rules were implemented with Drools, allowing fraud rules to be updated without recompiling the core code. The CI/CD pipeline in GitHub Actions validates and deploys new rules in a Blue/Green environment on Kubernetes, ensuring secure updates. Monitoring was performed using Prometheus and Grafana, creating dashboards that track "Rejected due to Limit" and "Rejected due to Fraud" rates, aligned with business metrics defined according to SRE practices.
  

Real-Time Credit Limit Management System

• I implemented the service using Quarkus with reactive endpoints via Mutiny, ensuring fast startup, low memory consumption, and immediate response. This improves pod efficiency in Kubernetes (EKS) and reduces costs. For limit storage, I used Caffeine as a local cache and AWS ElastiCache (Redis) as a distributed cache, applying an advanced cache-aside strategy. Redis was provisioned via Terraform, and the Java service includes fallback and failover in case the cache becomes unavailable. For integrity, I implemented the Two-Phase Commit pattern to maintain strict consistency between the limit service and the debit/credit service, meeting regulatory requirements for data integrity. For performance testing, I ran large-scale load tests with Gatling, integrated into the CI/CD pipeline in GitHub Actions, ensuring that performance issues are detected before production deployment.
  

Digital ticketing and sales system

• This project, a facial recognition identity validation system for Santa Cruz Futebol Clube, complies with the General Sports Law (14.597/2023) and promotes greater security, speed, and comfort for fans. The project integrates advanced AWS technologies such as Amazon Rekognition for facial recognition, DynamoDB for secure data storage, API Gateway, and AWS Lambda for serverless processing, ensuring high scalability and availability exceeding 99% on game days. Focusing on fraud reduction, elimination of ticket scalping, and faster access, the system offers rapid biometric authentication (less than 2 seconds), real-time notifications, and dashboards for access management and auditing. All this is done in full compliance with the LGPD (Brazilian General Data Protection Law), ensuring data encryption and rigorous permission control. I used the Scrum agile methodology to lead the development, implementing infrastructure as code with Terraform, containerization with Docker, and CI/CD pipelines for continuous and secure delivery. The entire scenario was designed exclusively to expedite the arrival of large crowds at stadiums, thus preventing congestion and long waiting times for fans to enter.