AI Specialist at edrone
I love Machine Learning. I am interested in: NLP, Reinforcement Learning , multiprocessing with python, development of custom neural models, PyTorch & TensorFlow library.
*** Employment History
NLP Engineer at Samsung R&D Institute Poland (02.2018 - 11.2020)
I had the pleasure to work in the AI Team, where I took part in many exciting projects in the field of NLP. These projects, among others, included: handwriting recognition, neural spellchecker, neural fact-checking systems (information retrieval & entailment with the Bert model), NLG with Generative Enhanced Model (custom modification of GPT-2 architecture).
AI Specialist at edrone (12.2020 - present)
I currently work in the company that deals, among other things, with an ecommerce chatbot development. We take care of research and development projects mainly related to information retrieval, natural language search, and recommendation systems. If you are interested in what we do, please check our blog: edrone blog.
While working at Samsung, I have participated (together with my colleagues from the TMLab team) in two international AI competitions:
1st prize, FEVER 2.0 (2019), Hong Kong
Our Generative Enhanced Model has been awarded the first prize on the FEVER 2.0 Breakers Task. I presented the solution at the ‘Fact Extraction and Verification’ workshop, which was a part of the EMNLP Conference.
2nd prize, SemEval (2019), Minneapolis
We took part in Task 8: Fact Checking in Community Question Answering Forums, Subtask A, and won the second prize. I participated in the NAACL conference, where I had a poster presentation of our method.
*** Private Projects
- python package with some little but frequently used python tools
ompr - object based multiprocessing tool, if your task needs a lot of processing power and it is possible to split it into subtasks, you may use this framework to do the job with the multiprocessing power
torchness - PyTorch tools, layers, encoders and much more
hpmser - hyperparameter search tool, features: simple management of parameters space, sampling optimization, multiprocessing, multi-GPU, TensorBoard visualisation, save, restart.
pypoks - Deep Reinforcement Learning for Poker with Python. Quite a big project dedicated to solving Poker games using Deep Reinforcement Learning; implemented from scratch with a lot of Neural Networks, Asynchronous Programming, Multiprocessing and Genetic Algorithms.
+48 501 245 737