Product: Pezconfiable. Peruvian artisanal-fishing startup with traceability. Sole owner of the product's technology direction. Lean team (1 backend, 1 frontend) using LLMs as a productivity multiplier. Own the product's overall architecture and all decisions on cloud platform, deployment topology and backend stack. Designed the backend on Java 21 + Quarkus and a Python oceanographic data processing module feeding traceability and fishing recommendations. Defined the AWS deployment topology optimized for the MVP stage (backend, PostgreSQL and Keycloak co-located on a single EC2; oceanographic processor on a dedicated EC2), with an explicit evolution plan once the product validates traction. Stack: Java 21, Quarkus, Python, PostgreSQL, Keycloak, AWS (EC2).

Client: Wabtec (USA) — PDS 2.0 (next generation of the Precision Dispatch System for railway network optimization). Approximately 20-microservice architecture; my team owns the track restrictions module. Designed an internal library for concurrent reservation of correlative identifiers; ships its own schema via Liquibase migrations packaged inside the Maven artifact. Adopted by 2 services in the product. Co-designed with the tech lead the parent-child state management of restrictions and the edit and copy flows; documented the solutions in design docs reviewed before being turned into user stories. Led the implementation of the restriction type with the most complex business logic in the module and refined the shared library to eliminate redundant database reads/writes. Actively contributed to end-to-end designs together with the tech lead, proposing alternatives and validating trade-offs. Stack: Java 21, Quarkus, Kafka, PostgreSQL on AWS RDS, AWS EKS, Liquibase, Jenkins + GitLab CI/CD, Grafana.

Client: Lulo Bank (Colombia, neobank). Back-processes team: nightly batch jobs closing the bank's accounting day (Spring Batch + AWS Lambdas in Python). Redesigned the TDA process from scratch (a Colombian banking regulatory obligation), working with non-technical domain experts; replaced a previous version carrying bugs and unstable patches. Diagnosed and fixed a numerical-precision defect in the IRR calculation (XIRR) caused by loss of convergence with Math.pow on double; replaced it with a library supporting BigDecimal-to-BigDecimal exponentiation. Owned evolutive maintenance of the main Spring Batch job and strengthened Python on the auxiliary Lambdas. Stack: Java, Spring Batch, Python, AWS (Lambda, SQS, EventBridge, RDS, CloudWatch, EKS), GitHub Actions.

Co-founded a 5-partner startup and took the CTO role. We sold the Peruvian government (INDECI) an offline-capable system for Damage Evaluation and Needs Analysis (EDAN) after disasters. Defined the backend architecture (Spring Boot 3-layer monolith on PostgreSQL), designed the data model and the REST API serving both the web portal and the field mobile app, and produced the technical documentation. Solved the client's on-premise requirement by designing a custom FTP service for ingesting photos from the mobile app in disaster zones. Led the initial deployment on GCP and later packaged the system for on-premise delivery (installer, APK, manuals, training sessions). Stack: Java, Spring Boot, Spring JDBC, MapStruct, PostgreSQL, FTP, Firebase Auth, Google Cloud Platform.

Migration of B2miningdata from monolith to microservices (B2B procurement: RFQs, quotations, orders, electronic invoicing). Assigned to the cross-cutting microservices group under the direct mentorship of the project architect. Designed and implemented the file management microservice (Azure Storage) and the centralized logging microservice on ELK, leading research, design and validation with the architect. Co-implemented the Event Router on Apache Kafka, the core piece of the product's Event Sourcing pattern, consumed by nearly all business microservices. Participated in design decisions with the architect, including the compute node network topology on Azure (AKS). Stack: Java, Spring Boot, Spring Cloud Netflix (Zuul), Apache Kafka, Azure (Storage, AKS), Docker, Kubernetes, ELK, PostgreSQL, MongoDB.