Ondřej Měkota

Machine Learning Engineer at Heureka.cz. I’ve read computational linguistics at Charles University in Prague.

You can find me on Github, LinkedIn or you can send me an email.

Experience

• August 2021 – now Heureka Group; Machine Learning Engineer
  Our team designs and runs a machine learning pipeline for offer/product matching. I work with neural networks, transformer language models, random forest models. We train and deploy using Gitlab CI/CD, Kubernetes, Google Cloud Platform, MLFlow, Neptune, DVC, Ray.
• December 2021 – now Czechitas; Lecturer
  Czechitas is non-profit organisation; I teach Python courses
• July 2020 – June 2021 SAP; Data Science Intern
  I worked on pattern mining in logs stored in an ElasticSearch cluster.
• October 2020 – February 2021 University of Jena; Contract Software Developer
  Short-term project. I developed a pipeline for morphological tagging, lemmatization and alignment in 30+ languages.

Education

• 2019–2021, Master’s (Mgr.) in Computational Linguistics
  Institute of Formal and Applied Linguistics, Faculty of Mathematics and Physics, Charles University. My thesis is on link prediction in inferred social network. I graduated summa cum laude.

• 2016–2019, Bachelor’s (Bc.) in Computer Science
  Faculty of Mathematics and Physics, Charles University. My thesis is on GAN-based anomaly detection.

Skills

• programming
  • Python (with ML, data packages)
  • I’ve written several blogposts about Python
  • some knowledge of C++, C, bash and Go
  • basic C#, PHP, and JavaScript skills
• machine learning
  • knowledge of recent deep learning methods
  • thorough knowledge of ‘common’ ML methods, unsupervised learning, graph neural networks
  • thorough knowledge of ML used for natural language processing (machine translation, NER, semantic/syntactic/morphological analysis): transformers, word embeddings, HMMs, CRFs,…
  • MLOps, hands on experience with deploying and maitaining large scale models in production
  • reinforcement learning
  • hands on experience with implementation of state-of-the-art ML methods, e.g. transformer, in PyTorch and Tensorflow
• databases
  • ElasticSearch, Redis, PostgreSQL
• devops
  • Docker, Kubernetes, Google Cloud, MLFlow, Neptune.ai, Gitlab CI/CD

Projects

• vim-python-docstring
  An open source plugin for Vim to automatically generate docstrings for Python source code.
  Downloaded by several thousands of users every year.