Jose Maria S.

I build the backend that keeps LLM features reliable in production — on top of 17 years of building distributed systems at PayPal/Braintree, Mercado Libre and Upwork. What I can do for you: • LLM backends: chat/agent APIs, streaming, tool calling, RAG, conversation state and concurrency • LLM evaluation: LLM-as-a-Judge pipelines and rubric scoring, so you know whether a prompt or model change made things better or worse • Java/Spring Boot microservices: new services, Java 21/25 upgrades, reactive APIs (WebFlux), performance and concurrency fixes • Payments & fintech: gateway integrations, idempotency, transactional consistency (Saga), PCI-aware design Where I've done it: • Upwork, Applied AI Platform: backend for Uma (Upwork's AI assistant), an LLM evaluation service and an agentic platform • PayPal/Braintree: payment gateway features and the GraphQL service for merchants • Naranja X: Tech Lead for core banking and prepaid card issuance • Mercado Libre: warehouse management APIs for Mercado Envíos Stack: Java 21/25, Spring Boot, Spring AI, Spring Batch, WebFlux, PostgreSQL, Redis, AWS, Docker, Kubernetes, GraphQL, Ruby on Rails. How I work: clear written updates, tests and observability by default, and good overlap with US time zones (Buenos Aires, UTC-3). Tell me what you're building and I'll reply with how I'd approach it.