← home
Michał Junczyk

Michał Junczyk

I turn real-world complexity into data for developing and evaluating AI systems.

Warsaw, Polandkontakt@michaljunczyk.plLinkedIn

Experience

  • Allegro, Warsaw12.2021 — present

    Head of Data Annotation & QA Services

    • Built the quality-assurance framework that gates AI products before release — definitions of done, acceptance criteria, and monitoring standards.
    • Scaled the organisation’s data-annotation capability — onboarded external vendor partners, rebuilt the annotation toolchain, and led the team that trained 700+ people across product and technology in Data-Centric AI over four years.
    • Deliver the datasets used to train and evaluate AI solutions across translation, search, the product catalogue, and many other areas — serving both end customers and teams inside the organisation.
    • Delivered 200+ annotation projects for AI teams across the organisation.
  • Samsung, Warsaw04.2017 — 11.2021

    Head of Language Data Operations

    • Led an end-to-end ML dataset operation for NLP systems — a Polish annotation team plus operational teams in Germany, France, Spain, and Italy (50+ engineers and domain experts across 5 countries).
    • Built and ran the data infrastructure and annotation toolchains the multinational R&D teams depended on for NLP dataset delivery.
  • Samsung, Warsaw03.2011 — 03.2017

    Data Engineer and Technical Leader

    • Led speech data collection and transcription pipelines for Automatic Speech Recognition across 8 languages and Samsung voice-enabled devices, feeding the voice assistants shipped in European markets.

Education

  • Adam Mickiewicz University, Poznań09.2019 — 06.2024

    Ph.D. in Computer Science

    Application of speech datasets management methods for the evaluation of Automatic Speech Recognition systems for Polish.

  • Poznań University of Technology10.2005 — 06.2010

    M.Sc. in Telecommunications

    Specialization: multimedia systems and signal processing.

Publications

Profiles: Google Scholar · ResearchGate · dblp

  • Junczyk, M., Christop, I., & Pęzik, P. (2024). Task 3: Polish Automatic Speech Recognition Challenge. PolEval 2024 Workshop. [pdf]
  • Junczyk, M. (2024). BIGOS V2 — Benchmark for Polish ASR: Curated Datasets and Tools for Reproducible Evaluation. NeurIPS 2024 (Datasets & Benchmarks Track). [poster]
  • Junczyk, M. (2024). A survey of Polish ASR speech datasets. Poznań Studies in Contemporary Linguistics 60(1), 27–52. [doi]
  • Junczyk, M. (2023). BIGOS — Benchmark Intended Grouping of Open Speech Corpora for Polish ASR. FedCSIS 2023, Warsaw, 585–590. [pdf]
  • Jeziorski, A., Sawicki, F., Solop, O., Junczyk, M., Sikora, M., & Zietkiewicz, T. (2020). Industrial ASR troubleshooting tool. LREC 2020 Industry Track, 10–14, ELRA. [pdf]

Talks

Teaching

  • AI Evaluation for Everyone — free online course for beginners and domain experts

    Open course · 2026 — present · course

  • Workshops on speech recognition systems evaluation

    AMU · 2024 — present · syllabus (full-time, part-time)

  • Python for Linguists (IT Skills for Linguists)

    AMU, English Philology dept. · 2022 — 2023 · materials

  • Introduction to Data Engineering

    AMU, Mathematics and Computer Science dept. · 2020

Projects

  • BIGOS V2 — Benchmark for Polish ASR

    Curated benchmark grouping 24 open Polish speech corpora (BIGOS V2 and PELCRA for BIGOS) into a single, reproducible evaluation dataset. Published at NeurIPS 2024 (Datasets & Benchmarks Track).

  • Polish ASR Leaderboard

    Live public leaderboard ranking 25 speech-to-text systems on 4000+ Polish recordings — accuracy, cost, speed, and robustness.

  • Polish ASR speech datasets catalog

    Open catalog of speech datasets for Polish, intended for ASR systems development.

Community & service

  • GenAI in Localization 20262026

    Organizing committee [committee]

Skills

  • Data-Centric ML systems design and development
  • Data Annotation
  • Data Engineering
  • Automatic Speech Recognition
  • Team management
  • Project management

Awards

Best R&D Team Leader 2018 — recognized by the head of Samsung Research Institute Poland for contributions to the commercialization of voice assistants in European markets.

Languages

English, Polish