Asier Cabello - R&D Engineer

~/cv/ASIER_CABELLO_en.txt
UTF-8 • LF • mono

Curriculum Vitae Editor View

▸ editor ASIER_CABELLO_en.txt read-only
// ===============================================================
// ---------------------- CURRICULUM  VITAE ----------------------
// ===============================================================

ASIER CABELLO
Location: Bilbao, Bizkaia
Phone: +34 688 86 73 97
Email: asier.cabello.s@gmail.com
Website: asiercabello.dev
GitHub: github.com/Txoka
Hugging Face: huggingface.co/Txoka
LinkedIn: linkedin.com/in/asier-cabello

/* ABOUT */
AboutMe:
  R&D Engineer specialising in music information retrieval, neural audio embeddings,
  similarity search and applied machine learning. Experienced in taking research ideas from
  experimental design and benchmarking through to production systems, including music identification
  pipelines, large-scale retrieval infrastructure and LLM-assisted automation. Multidisciplinary by
  nature, I learn new domains quickly and apply the same research rigour and pragmatic engineering
  across different technical fields.

/* EXPERIENCE */
Experience:
  2025 - Present  R&D Engineer  @ BMAT
    - Worked on core R&D initiatives across music identification, including cover detection,
      neural music fingerprinting, AI-generated music detection, music metadata matching,
      LLM-assisted validation workflows, audio embedding retrieval, and dataset/benchmark design.
    - Modernized the fingerprint matching stack by refactoring the infrastructure and delivering a
      production-ready embedding-based identification pipeline using faiss.
    - Re-architected and refactored one of the company’s core technologies, delivering 16% higher
      saturated throughput, reducing p99 latency to ~1/3 at low load and ~1/5 at high load,
      and reducing memory usage to 1/4–1/5, using half as many workers.
    - Created and led an internal chatbot project leveraging multiple internal data sources;
      managed a small team and drove execution end-to-end.
    - Developed an AI-agent based automated QA system using OpenAI API that outperformed human
      reviewers at a fraction of the time and cost and transformed internal processes company-wide.

  2022 - 2024  Developer / Data Scientist  @ Azterlan
    - Developed .NET applications with AI features.
    - Built computer vision models for metallography analysis.
    - Trained predictive models for metal quality.

  2020 - 2024  Private Programming Tutor
    - Programming + algorithms lessons to university students.

/* EDUCATION */
Education:
  2020 - 2024  Degree in Data Science and Artificial Intelligence
               University of Deusto
               - Honors in Final Degree Project
               - Extraordinary Degree Award: best academic record in the cohort

  2018 - 2020  Technological High School
               Colegio Santa María, Portugalete

/* PUBLICATIONS */
Publications:
  AI-Generated Music Detection in Broadcast Monitoring
    - Co-authored an ICASSP paper introducing AI-OpenBMAT, the first public dataset designed
      for AI-generated music detection in broadcast monitoring scenarios.
    - Helped shape dataset and experimental design decisions and implemented one of the baseline
      detection systems.

/* PERSONAL PROJECTS */
Projects:
  Training Reasoning LLMs
    - Fine-tuned an instruction-tuned Llama 3B model using GRPO and a labeled dataset
      of mathematical problems leveraging curriculum reinforcement learning.

  Service Development and Homelab Deployment on Linux
    - Built web apps, APIs, and services; Linux server management.
    - Used Docker, Apache, Nginx, FastAPI, React.

  Fine-tuning Language Models
    - Generated/selected/filtered synthetic data for augmentation.
    - Fine-tuned and evaluated Mixtral as an assistant; served through an API.

// ;)