Cofounder

Caudals

Cofounded an applied AI company specializing in enterprise AI evaluation and custom dataset engineering. Architected the platform to benchmark LLMs, assistants, and autonomous agents against golden test sets derived from proprietary documentation.

Engineered automated evaluation pipelines, scoring engines, and regression-tracking workflows for recurring subscriptions. Built the data infrastructure that translates diagnostic findings into high-quality custom datasets created with a managed network of freelance domain experts.

AI Engineer Intern

Altia

Architected an AI-driven automation framework leveraging autonomous agents and specialized skills to parse regulatory specifications within the CAE platform.

Engineered an end-to-end multi-circuit validation suite combining browser automation with Playwright and REST APIs to run positive and negative synthetic tests, and real-case testing across headless CI/CD pipelines.

Research Intern - Telematics and Image Group

Department of Signal Theory, Communications and Telematics Engineering, University of Valladolid

Researched computer vision and deep learning for biomechanical movement analysis. Benchmarked 3D pose estimation models against inertial sensor references, publishing two first-author journal papers.

Engineered video-IMU signal synchronization pipelines and trained BiLSTM recurrent networks to automatically segment gait phases from monocular video for telerehabilitation.

MSc in Telecommunications Engineering

ETSIT, University of Valladolid

BSc in Telecommunications Engineering

ETSIT, University of Valladolid

42 Common Core

42 School, Telefonica Madrid

Baccalaureat general, scientific specialisations

Clermont-Ferrand Academy

Caudals - B2B AI Dataset Marketplace

Building a startup that helps companies monetize proprietary data and helps AI teams buy clean, compliant, ML-ready datasets for training, fine-tuning, evaluation, and enrichment.

The product combines buyer and supplier intake, dataset feasibility studies, rights and privacy review, preprocessing, cleaning, anonymization, enrichment, labeling, QA, documentation, and delivery through private offers or future marketplace listings.

Connected Vehicle Prototype - 4G, AI, and Autonomous Control

Project Workshop II prototype integrating a 4G/LTE network, artificial intelligence, and a connected vehicle workflow to detect traffic signs and act on a small autonomous car.

Worked from a system architecture using srsRAN 4G, BladeRF, MQTT, AWS DeepRacer, Roboflow, YOLO26, Jetson Orin Nano, LTE performance measurements, and edge inference to evaluate latency, reliability, and real-time control constraints.

IoT Datalogger - Electronic Systems

Designed and integrated a radiosonde-inspired IoT datalogger with four stacked PCBs: power and programming, FPGA-based control, environmental sensing and GPS, and LoRa communications at 868 MHz. The prototype used Proteus for schematic capture, custom footprints, PCB routing, Gerber generation, BOM management, and fabrication files.

Worked on a system including a Lattice iCE40HX4K FPGA, a RISC-V control architecture, C firmware, Verilog peripherals, UART/SPI/I2C sensor interfaces, BME680 environmental sensing, GPS, MQ9 gas sensing, GP2Y1010 dust sensing, and LoRa data transmission.

IoT Datalogger - Communication Systems

Designed, simulated, fabricated, and validated the RF communication subsystem for the datalogger. The system used LoRa at 868 MHz through an SX1262-based module, SPI-controlled prototyping boards, and a base station workflow for environmental and GPS measurements.

Compared an external patch antenna and a PCB-integrated folded monopole antenna using ANSYS HFSS, NanoVNA measurements, S11, impedance, VSWR, link budget calculations, Xirio Online coverage studies, and lab transmission tests.

Deep Learning for Human Pose Estimation and Gait Segmentation

Evaluated MotionAGFormer, MotionBERT, VideoPose3D/MMPose, and NVIDIA BodyTrack for 3D human pose estimation from monocular video against IMU reference data. MotionAGFormer achieved the strongest global performance in the evaluated dataset, with a 9.27 degree RMSE and a 0.86 Pearson correlation.

Developed a gait phase segmentation workflow using IMU angular velocity, supervised labelling, and a BiLSTM model trained on knee joint angle sequences. The model reached more than 80% F1-score and accuracy on inertial data, and around 70% on video-derived angles.

IBM Data Analyst Specialization

IBM

Machine Learning Specialization

DeepLearning.AI

Fundamentals of Deep Learning

NVIDIA

Introduction to Deep Learning with PyTorch

UVaIA Center

Computer Vision Certification

UVaIA Center

Digital Competences Certificate B2

Regional Government of Castile and Leon

German B1 Certificate

University of Valladolid Foundation

Driving Licence B

Spanish Directorate-General for Traffic

Certificate of Proficiency in English C2

University of Cambridge

DELF B2 French Diploma

Alliance Francaise

Santander X Spain Awards - University Entrepreneur Projects

Banco Santander

First Prize T-CUE Campus Emprendedor

FuesCyL, JCyL

First Prize TalenTIC 2025 for ICT innovation

Inetum, University of Valladolid

Consolida Startup Program

IdeVa, Valladolid City Council

Extraordinary Baccalaureate Award

Regional Government of Castile and Leon

Extraordinary Secondary Education Award

Regional Government of Castile and Leon