Skip to content
Roque Sobrino
ES Email me

Roque Sobrino

AI Engineer · Vigo, Spain

roquesobrino.pages.dev · roquesobrino@gmail.com · github.com/roquesobrino · linkedin.com/in/roque-sobrino-lago

Download as PDF ↓

AI engineer building the AI department of FERRI, a marine machinery manufacturer, from scratch: RAG over the sales archive, BI on a legacy ERP and CAD→ERP automation — all of it in production, on-premise and verified. First Class graduate in Artificial Intelligence.

Experience

AI Engineer (IT department)

Sep 2025 — present

FERRI (Industrias Ferri S.A.) · Vigo

  • RAG assistant over decades of sales proposals (LangGraph + Neo4j, on-premise), in pilot production with a deterministic evaluation of 53/53.
  • BI dashboard on a legacy ERP of ~1,460 tables: full reverse engineering, delivered 4 days ahead of the deadline.
  • Trilingual corporate website (ferri-sa.es) rebuilt in-house with Astro, launched without a second of downtime.
  • Bills of materials from CAD to the ERP: 95.7% of lines exact across 210 machines validated against the real ERP.
  • Parametric generation of DIN fasteners in SolidWorks: 7 families and ~44 parts verified.

Warehouse and hospitality jobs

Oct 2021 — Nov 2023

Leicester (UK) · Vigo

  • I paid for my university years by working while studying the degree full time.

Education

BSc (Hons) Artificial Intelligence — First Class

Sep 2022 — Jul 2025

De Montfort University · Leicester (UK)

  • Overall average of 74; final project (Development Project) with 86 — 100/100 in the project component.
  • Dissertation: fish freshness from images with transfer learning (98.19% test accuracy, EfficientNetB0).
  • Modules: Big Data & Machine Learning, Agent-Based Modelling & Parallel Computing, Applied AI, Intelligent Robotics, Fuzzy Logic.

International Year 0 — English and Computer Science

Sep 2020 — Jun 2021

De Montfort University · Leicester (UK)

Spanish Bachillerato, science and technology track

Sep 2017 — Jun 2020

Compañía de María · Vigo

Certifications: Generative AI (MIT Professional Education), Santander Open Academy (2024) · Crash Course on Python (Google), Coursera (2025).

Academic AI projects

  • Fish freshness classification with CNNs (dissertation) — transfer learning and a custom hybrid model; 98.19% test accuracy
  • Classification of 5 rice varieties with CNNs — 98% with ResNet50, comparing four pretrained networks with transfer learning
  • Robot localisation with an Extended Kalman Filter — real-time sensor fusion
  • Horseshoe wear prediction with deep learning — proactive maintenance from usage data

Stack

AI and data
PyTorch / TensorFlow · LangGraph · Vertex AI (Gemini) · RAG · Vision (CNN, OpenCV) · Neo4j · SQL Server / PostgreSQL
Backend
Python · FastAPI · TypeScript · REST · Zod
Web
Astro · React · Tailwind CSS · Three.js · GSAP
Operations
Docker · nginx / Traefik · Coolify · Git/GitHub · n8n

Languages: Spanish (native) · English (fluent — 4 years living and studying in the UK).