Marcos Javier Alvarez Maestre

Core engineer on a 15-service TypeScript monorepo. Designed and shipped LLM pipelines for document data extraction, taking accuracy from 30% to 87% — at least as reliable as the incumbent OCR vendor, with wider coverage, at about 3% of its cost. Built email/SMS infra, LLM classification, and agentic workflows with eval harnesses and cost tracing.

Took over the client relationship mid-project after the PM left: diagnosed a roadmap built on misread requirements, renegotiated scope and deadlines with a frustrated client, and restored the account — which won the company further work.

Migrated quote data to the rebuilt platform and reconciled it against the legacy system; traced a mismatch in financial totals to differing rounding behavior and aligned the new calculations to the client's existing rules.

Led the architecture and the development team for a serverless platform that generates, compiles and manages dynamic documents for a major supply-chain management provider. Owned infrastructure as code and automated deployments, plus R&D across backend and frontend.

Migrated a self-hosted monolith to an AWS Kubernetes microservices cluster; migrated a mobile backend off a discontinued BaaS ahead of its shutdown, reverse-engineering undocumented internals from source.

Full-stack and DevOps delivery for IoT platforms that broadcast academic content, tests and video conferences from web and mobile apps to embedded devices, plus a video-intercom product with remote door control. Owned deployments and R&D on experimental hardware.

Built a high-performance backend for a visual-novel game with micropayments, and an e-learning platform for course and content delivery. Automated deployments, integrated third-party services and worked across backend and frontend.

Part-time undergraduate studies alongside a full-time engineering career, focused on software engineering fundamentals, distributed systems and data.