08/24/2026 updated


100 % available
Senior Backend & AI Engineer · Java/Quarkus · MCP servers, agent frameworks · GraalVM native
Murcia, Spain
Worldwide
Associate Degree of Software EngineeringAI AgentsAWS SQSRAGGitHub CopilotGitHub ActionsTailwind CSSOpenTelemetryLLMJava (Programming Language)JavaScriptAmazon Web ServicesAmazon S3Atlassian JiraGoogle BigQueryConcurrent Computing
Senior backend and AI engineer with 5+ years delivering production systems end to end: Java/Quarkus services, Next.js frontends and the AI agent infrastructure that runs them - MCP (Model Context Protocol) servers, multi-language agent frameworks and evaluation harnesses, not agent demos. Measured results shipped in the last 12 months include +58% sustained platform capacity (330 to 520 rps), -45% p95 latency and -98% serverless invocations on live products. Previously core engineer on a distributed, high-concurrency platform serving 15M+ users, where I cut push notification delivery from 50ms to 1ms on average. Currently CTO and Tech Lead for two product companies, owning architecture decisions and defining how success is measured before work starts. Registered self-employed in Spain, working remotely with EU and US clients on senior contract engagements: 0 to 1 product delivery, LLM and agent integration into real products, and performance rescue on Java backends.
Experience
Professional experience5 - 10 years
Freelancer since2026
Languages
EnglishFluentSpanishNative speaker
Project history
Independent product studio, Murcia, Spain.
- Consolidated 10 applications and 99 modules (78 Maven, 21 pnpm) into a Turborepo/pnpm monorepo, unifying 307 REST endpoints under a shared build chain and removing duplicate React copies that caused runtime failures.
- Designed seven domain-segmented CI/CD pipelines with real PostgreSQL and Redis in the integration suite and reproducible artifact publishing from CI rather than developer machines.
- Architected a programmatic SEO system serving approximately 68,000 routes at constant build cost (794 prerendered pages plus on-demand ISR), reaching Total Blocking Time of 0-10 ms and Cumulative Layout Shift of 0.
- Designed an MCP server tool surface that does not scale with the number of sources (4 constant tools, 55% of the context budget), selecting between two architectures by measurement after proving the alternative would publish 47 tools and fail to start at 45 sources.
- Cut serverless invocations by 98% and billed compute by 75% on a real-time audio product by collapsing requests at the CDN and right-sizing per-function memory.
- Build multi-language agent frameworks (TypeScript, Java, Python), visual tooling for agentic flows, and systems that turn expert judgment into reusable AI artifacts.
Stack: Java 25, Quarkus, GraalVM native, Next.js 16, React 19, TypeScript, Python/FastAPI, PostgreSQL, Docker, GitHub Actions, MCP.
Digital legal services platform on Java/Quarkus and Next.js, Murcia, Spain (remote).
- Led the migration of a 103,000-line Next.js application from Pages Router to App Router while simultaneously introducing the Playwright end-to-end (E2E) suite the project did not previously have.
- Made the E2E suite 5.2x faster (4m 12s to 48s) by redesigning the navigation pattern of a single spec that accounted for 31% of total suite time, preserving 100% of assertions.
- Redesigned the E2E quality gate and removed approximately 8 minutes of CI per pull request: execution moved to the environment where real dependencies exist, with CI reading an attestation bound to the commit SHA that blocks merges when not green.
- Diagnosed two production regressions invisible to the build pipeline after a Quarkus REST engine migration, including an endpoint returning a permanent 500 because a reactive return type pulled the entire filter chain onto the event loop; verified across 253 endpoint/method pairs with real authentication.
- Built flow orchestration and automated cross-referencing of public data sources (official gazette, company registries) for business-opportunity detection: 285 REST endpoints, 66 entities and 108 versioned database migrations.
Stack: Java 21, Quarkus, Next.js 16, React 19, Tailwind 4, PostgreSQL, Liquibase, Playwright.
High-volume transactional platform serving 15M+ registered users across Spain, Mexico and Colombia, with peaks in the thousands of concurrent users; core development and architecture across 22+ repositories inside a backend organization of about 40 engineers.
- Worked on a Quarkus SPEI payments microservice handling approximately 40k transactions per day, and integrated a banking provider (Bancomer) from scratch, building custom fixed-position file mappers to their specification, with strict position handling and an event-based rollback mechanism for transaction recovery.
- Ran the programmatic SEO / GEO initiative reporting directly to the CEO, and proposed the AI delivery workflows that became the company's starting point for using AI in engineering.
- Delivered the end-to-end technical launch of the Colombian market (backend, CRM, microservices) and built core services for the Mexican market that were designed to apply across all countries.
- Analysed Product/UX requirements directly, wrote the technical cards for backend and frontend teams, and took architectural decisions on complex business flows.
- Applied distributed and local Redis caching and managed transactionality; reviewed other engineers' pull requests as a quality gate; shipped production deploys and hotfixes on 24/7 on-call rotations (L3 and L4).
Stack: Java 17, Jakarta EE, Spring Boot, Quarkus, WildFly, React, Next.js, AngularJS, PostgreSQL, RabbitMQ, Redis, FCM, AWS (Lambda, S3), CDN, Jira.