Portrait of Liam Düro, smiling, arms crossed, in a black hoodie against a pale wall.

Liam Düro

Machine learning engineer & data scientist

Now Currently: Machine Learning Engineering Intern at BMW Group, Munich.

I build and evaluate machine learning systems, most recently retrieval-augmented generation for telecom. I have just finished an MSc in Data Science and Machine Learning at UCL and am now a machine learning engineering intern at BMW Group in Munich.

Experience

  • BMW Group Machine Learning Engineering Intern, Munich Oct 2026 – Mar 2027
  • Huawei Paris Research Center Postgraduate Researcher, Paris Feb 2026 – Sep 2026
  • Tuomi Data Science and Backend Development Intern, Luxembourg May 2025 – Jul 2025

Where I've worked and studied

World map with two zoom areas Simplified outline of the world. Two boxes mark the areas shown in more detail below: Europe, and Australia. Five small dots mark places: four in Europe and one in Sydney, Australia. Paris (remote work) appears only in the zoomed Europe panel. Europe Australia
Europe, zoomed in Simplified outline of Ireland, Britain and western and central Europe, zoomed in from the world map, showing Dublin (study), London (study), Paris (work, remote), Luxembourg (work) and Munich (work, current). Europe DublinTrinity College Dublin LondonUCL Paris (remote)Huawei Paris Research Center (remote) LuxembourgTuomi MunichBMW Group (now)
Australia, zoomed in Simplified outline of mainland Australia and Tasmania, zoomed in from the world map, with one pin: Sydney (study, exchange year 2023 to 2024). Australia SydneyUniversity of SydneyExchange year, 2023–24
Simplified outlines from Natural Earth, not to scale.
  • Work
  • Work, remote
  • Study
  • Now
  • Dublin Trinity College Dublin, study
  • Sydney University of Sydney, exchange year 2023–2024, study
  • Luxembourg Tuomi, work
  • London UCL, study
  • Paris Huawei Paris Research Center, work (remote), Feb to Sep 2026
  • Munich BMW Group, work, now

Selected work

  1. A RAG benchmark for 3GPP telecom specifications

    A retrieval-augmented generation pipeline and benchmark for evaluating large language models on 3GPP specifications, built at Huawei Paris Research Center as part of the GSMA Open Telco AI Initiative. GSMA adopted the benchmark as the basis for an industry competition.

  2. Energy consumption of a virtualised 5G network

    My final-year project at Trinity: a Python data pipeline that streams metrics from a virtualised 5G network (srsRAN and Open5GS) through Telegraf into InfluxDB, then uses machine learning to predict CPU energy use. A random forest model reached R² 0.94.

  3. AI Prompt Manager

    A Chrome extension (Manifest V3) for keeping a library of prompt templates and inserting them into ChatGPT and other AI chat boxes in one click. Listed on the Chrome Web Store as Prompt Inserter.

    Source on GitHub

Education

  • University College London MSc Data Science & Machine Learning 2025 – 2026
  • Trinity College Dublin BA Computer Science & Business Studies, First Class Honours 2021 – 2025