I am a Data Science student, mainly interested in developing Deep Neural Network-based Text-to-Speech systems. In the moment, I am attending the master's degree course in Computer Science and Intelligent Systems on AGH University of Krakow.
Feel free to check out my publication titled Using Denoising Diffusion Model for Predicting Global Style Tokens in an Expressive Text-to-Speech System that appeared in a special issue of Electronics. It contains results of months of research and work on developing a Deep Learning-based TTS system.
As for the professional experience, I've been able to gain a fair level of proficiency in C++ Software Development while working at Aptiv in Krakow. This gave me an opportunity to gain practical skills in developing object-oriented infrastructures, designing and implementing tests as well as optimizing and debugging the code.
I'm interested in applying advanced techniques for learning languages and currently I'm striving for getting an advanced level of fluency in German.
| Name | Description | Tech stack |
|---|---|---|
| ddpm-gst-speech-synthesis | A Deep Learning system performing Text-to-Speech synthesis with explicit prosody modeling based on diffusion generative algorithm. The project scopes data preprocessing, model training and evaluation of the system, both objective and human-based. | Python, Pytorch |
| (WIP) paragraph-tts | An advanced TTS system for synthesizing long-form speech, e.g. audiobook paragraphs. It scopes preparing complex numerical representations for large amounts of pre-processed data and training a Deep Learning model, leveraging advanced fast attention mechanism for improved length generalization. | Python, Pytorch, MLFlow |
| (WIP) ai_projects | A C++ framework that helps developing deep learning models. It contains building blocks for designing neural networks as well as significant algorithms for training it, i.e. back-propagation and tensor operations. | C++, Googletest |
| advanced-data-mining-project | Contains results of investigating sentiment analysis possibilities for a scraped dataset containing textual restaurant reviews. The repository contains pipeline for data preprocessing, involving composing advanced numerical representations for text content, and benchmarking multiple rating prediction models. The project includes professional experiment tracking and automated model's evaluation. | Python, Pytorch, MLFlow, Transformers |
| (WIP) comp-int-proj-tts | Contains implementation of a Conformer-based TTS model with several prosody modeling methods. The purpose of the project is to benchmark different configurations of the model in order to gain insight into the influence of the prosody modeling techniques on the model's performance. | Python, Pytorch, MLFlow |
| (WIP) rag-app | A containerized infrastructure serving an AI Chat that provides support on behalf of a chosen institution. The system spans a demo web application, a RAG-style knowledge base handling and a module responsible for providing content safety guardrails and streaming LLM responses. The infrastructure is designed to support self-hosted models. | Python, LangChain, Docker, FastAPI, Gradio, Qdrant |