Maria is an outstanding NLP Engineer and Data Scientist. She has an in-depth understanding of the latest NLP approaches and can apply them to actual industry problems. Specifically, Maria has an expert-level understanding of Transformer based neural networks. In addition to that, Maria is proficient with Python and PyTorch. She also has worked extensively with PySpark and other big data frameworks. As a colleague, Maria was always a team player and put a great emphasis on knowledge sharing and collaboration. Moreover, Maria always strived to become a better engineer as well as a team member, through consistently seeking feedback from her peers. Furthermore, Maria showed great skills in managing various team projects and being accountable for multiple team initiatives. For instance, she has supervised Thesis students, as well as, interns in our NLP team. She will be an invaluable asset to any software engineering team.

Maria did her master thesis "Data-Driven Pronunciation Generation for ASR" in our company Sony Europe B.V. from 01.02.2019 until 31.07.2019. During this time i was acting as her supervisor within the company. In ASR systems dictionaries are used to describe the pronunciations of all words of a language. The dictionary translates each word into a sequence of phonemes. Such dictionaries are typically hand crafted by a linguist expert. In this thesis we aimed at a data-driven approach for generating pronunciations for a dictionary. Pronunciation generation from data is considered a quite difficult problem and hand crafted dictionaries are still state of the art. During her thesis Maria implemented different methods for generating pronunciations for words from existing audio samples. In particular different pronunciation clustering methods have been researched and compared with existing baselines. In this context she used Neural Networks to learn phoneme characteristics. Maria has a very good linguistic and statistical background which was required for the project. She also showed to have the required good programming skills to solve the given tasks. She did a Python implementation of all the methods researched, basically from scratch. She also dealed with the complexity of the related Kaldi ASR toolkit in order to access the data and models required for her work. She always had a very structured and disciplined work style. Due to the difficulty of the task we had many discussions and she had always very creative ideas for solving the problems. She is a highly motivated researcher and she is always moving directly towards the goals. She has very good communication skills and it was a pleasure to work with her.

Maria is an outstanding NLP Engineer and Data Scientist. She has an in-depth understanding of the latest NLP approaches and can apply them to actual industry problems. Specifically, Maria has an expert-level understanding of Transformer based neural networks. In addition to that, Maria is proficient with Python and PyTorch. She also has worked extensively with PySpark and other big data frameworks. As a colleague, Maria was always a team player and put a great emphasis on knowledge sharing and collaboration. Moreover, Maria always strived to become a better engineer as well as a team member, through consistently seeking feedback from her peers. Furthermore, Maria showed great skills in managing various team projects and being accountable for multiple team initiatives. For instance, she has supervised Thesis students, as well as, interns in our NLP team. She will be an invaluable asset to any software engineering team.

Maria did her master thesis "Data-Driven Pronunciation Generation for ASR" in our company Sony Europe B.V. from 01.02.2019 until 31.07.2019. During this time i was acting as her supervisor within the company. In ASR systems dictionaries are used to describe the pronunciations of all words of a language. The dictionary translates each word into a sequence of phonemes. Such dictionaries are typically hand crafted by a linguist expert. In this thesis we aimed at a data-driven approach for generating pronunciations for a dictionary. Pronunciation generation from data is considered a quite difficult problem and hand crafted dictionaries are still state of the art. During her thesis Maria implemented different methods for generating pronunciations for words from existing audio samples. In particular different pronunciation clustering methods have been researched and compared with existing baselines. In this context she used Neural Networks to learn phoneme characteristics. Maria has a very good linguistic and statistical background which was required for the project. She also showed to have the required good programming skills to solve the given tasks. She did a Python implementation of all the methods researched, basically from scratch. She also dealed with the complexity of the related Kaldi ASR toolkit in order to access the data and models required for her work. She always had a very structured and disciplined work style. Due to the difficulty of the task we had many discussions and she had always very creative ideas for solving the problems. She is a highly motivated researcher and she is always moving directly towards the goals. She has very good communication skills and it was a pleasure to work with her.

Maria is an outstanding NLP Engineer and Data Scientist. She has an in-depth understanding of the latest NLP approaches and can apply them to actual industry problems. Specifically, Maria has an expert-level understanding of Transformer based neural networks. In addition to that, Maria is proficient with Python and PyTorch. She also has worked extensively with PySpark and other big data frameworks. As a colleague, Maria was always a team player and put a great emphasis on knowledge sharing and collaboration. Moreover, Maria always strived to become a better engineer as well as a team member, through consistently seeking feedback from her peers. Furthermore, Maria showed great skills in managing various team projects and being accountable for multiple team initiatives. For instance, she has supervised Thesis students, as well as, interns in our NLP team. She will be an invaluable asset to any software engineering team.

Maria did her master thesis "Data-Driven Pronunciation Generation for ASR" in our company Sony Europe B.V. from 01.02.2019 until 31.07.2019. During this time i was acting as her supervisor within the company. In ASR systems dictionaries are used to describe the pronunciations of all words of a language. The dictionary translates each word into a sequence of phonemes. Such dictionaries are typically hand crafted by a linguist expert. In this thesis we aimed at a data-driven approach for generating pronunciations for a dictionary. Pronunciation generation from data is considered a quite difficult problem and hand crafted dictionaries are still state of the art. During her thesis Maria implemented different methods for generating pronunciations for words from existing audio samples. In particular different pronunciation clustering methods have been researched and compared with existing baselines. In this context she used Neural Networks to learn phoneme characteristics. Maria has a very good linguistic and statistical background which was required for the project. She also showed to have the required good programming skills to solve the given tasks. She did a Python implementation of all the methods researched, basically from scratch. She also dealed with the complexity of the related Kaldi ASR toolkit in order to access the data and models required for her work. She always had a very structured and disciplined work style. Due to the difficulty of the task we had many discussions and she had always very creative ideas for solving the problems. She is a highly motivated researcher and she is always moving directly towards the goals. She has very good communication skills and it was a pleasure to work with her.