Kyrylo Vasyliev

Self-motivated problem solver with expertise in AI, specializing in Computer Vision and Natural Language Processing. Proficient in TensorFlow, NumPy, Pandas, Matplotlib, and Scikit-learn. Strong organizational and communication skills with a proven track record of placing top 3 in regional programming and math olympiads during high school. Fluent in English (C1 level) and adept at collaborating in a team environment.

• Beetroot Academy Python course, https://shorturl.at/yBccY
• Regional olympiads diplomas, https://shorturl.at/IMAh0, https://shorturl.at/VJ5ri

Face Recognition Model (March 2024 - April 2024)

• Developed a face recognition model using Conv2D layers and residual units on the DigiFace-1M dataset.
• Achieved 80% validation accuracy.
• Stack: Python, TensorFlow, Keras, Numpy, Pandas, Matplotlib

Neural Machine Translation Model (May 2024 - July 2024)

• Developed a model for translating English text into Spanish using transformer encoder-decoder architecture and BPE tokenization with 10000 merges.
• Achieved ~84% validation accuracy.
• Stack: Python, TensorFlow, Keras, Numpy, Matplotlib, Huggingface

RNN Model for Predicting Bach Chords (April 2024 - April 2024)

• Developed a model trained on the Bach chorales dataset to predict the next chords.
• Achieved 86% validation accuracy
• Achieved 58% CHRF for 3-grams
• Stack: Python, TensorFlow, Keras, MIDIFile, Matplotlib, Numpy

Telegram Bot for Music Quizzes (July 2024 - Present)

• Created a Telegram bot using aiogram and SQL for generating music quizzes.
• Increased Llama accuracy on creating quizzes from ~60% to 85% by utilizing RAG
• Stack: Python, Aiogram, MySQL, Redis, AWS Sagemaker, AWS S3, Docker, Weaviate, Elastic Search