~/adam.kaczmarek
Adam Jan Kaczmarek
// résumé

Adam Jan Kaczmarek

Deep Learning Engineer · NLP Researcher · Open-source Contributor

PhD candidate and senior engineer with a decade of applied research in natural language processing and computer vision. I co-organize PolEval, contribute to the CLARIN-PL open-source stack, and build production NLP systems for low-resource languages — from multilingual NER and coreference to ASR rescoring and information extraction. Equally at home in a paper draft, a CUDA profile, and a Kubernetes manifest.

adam.jan.kaczmarek@softwaremill.com+48 661 105 014githubacl anthology
// academic background

Education

  1. PhD candidate, Computer Science

    2017 – present
    Wrocław University of Science and Technology

    Research on neural NLP for low-resource Slavic languages; named entity recognition, coreference, and ASR rescoring.

  2. MSc, Computer Science

    2015 – 2017
    Wrocław University of Science and Technology

    Thesis on deep learning for sequence labeling. Graduated with distinction.

  3. BSc, Computer Science

    2011 – 2015
    Wrocław University of Science and Technology

    Specialization in algorithms and intelligent systems.

// career

Experience

  1. Senior Deep Learning Engineer

    2022 – present
    SoftwareML

    Lead NLP and applied DL initiatives for enterprise clients — model training, evaluation pipelines, and production deployment on GPU clusters.

  2. Research Engineer

    2018 – 2022
    CLARIN-PL (Wrocław Tech)

    Built and maintained open-source NLP tooling: Liner2, Inforex annotation platform, NER and coreference resolution models for Polish.

  3. Teaching & Research Assistant

    2017 – 2022
    Wrocław University of Science and Technology

    Taught ML, algorithms and software engineering; supervised BSc/MSc theses in NLP.

  4. Computer Vision Researcher

    2012 – 2017
    BioVision Lab

    Developed CellStar — segmentation and tracking of budding yeast cells in time-lapse microscopy.

// selected work

Projects

PolEval 2018 / 2020 / 2021

Co-organizer and contestant — NER, ASR rescoring, punctuation restoration shared tasks for Polish.

BSNLP Shared Tasks (2019, 2021)

Multilingual Slavic NER, entity linking, and coreference resolution submissions and tooling.

Liner2 & Inforex

Long-running contributions to CLARIN-PL open-source NLP stack used by research groups across Europe.

t-REx for ASR

Transformer-based rescorer + lattice extender; SoTA on PolEval 2020 ASR task.

CellStar & Yeast Image Toolkit

Cell segmentation algorithm and benchmarking platform for brightfield microscopy.

// talks & events

Public appearances

  • PolEval Workshop
    Co-organizer and presenter — punctuation restoration task overview.
    2021
  • BSNLP @ EACL
    Talk: Slavic NER submission and few-shot transfer.
    2021
  • PolEval Workshop
    Presentation: t-REx — transformer rescorer for hybrid ASR.
    2020
  • PolEval Workshop
    Talk: Nested Named Entity Recognition for Polish.
    2018
  • CLARIN-PL Tech Meetups
    Recurring talks on annotation tooling and NER pipelines.
    2019 – 2022
// papers

Publications

  1. 01
    Slavic Named Entity Recognition: BSNLP 2021 Shared Task Submission
    BSNLP @ EACL 2021
    ↗
  2. 02
    PolEval 2021 Task 1: Punctuation Restoration from Read Text
    PolEval 2021
    ↗
  3. 03
    t-REx: Transformer Rescorer & Extender for ASR
    PolEval 2020
    ↗
  4. 04
    Nested Named Entity Recognition for Polish
    PolEval 2018
    ↗
  5. 05
    CellStar: Algorithm for Yeast Cell Segmentation in Brightfield Microscopy
    Bioinformatics
    ↗
// toolbox & credentials

Skills & certificates

Languages
PythonScalaJavaC++BashSQL
ML / DL
PyTorchTensorFlowHuggingFacescikit-learnONNXCUDA
NLP
TransformersELECTRABERTNERCoreferenceASRTokenization
Infrastructure
DockerKubernetesMLflowAirflowAWSSpark