alex voronkov

Designed scalable medallion architecture for high-frequency trading data and developed transformer-based models for portfolio optimization. Increased 3-month trader retention by 15%.

Fine-tuned DinoV2 for skin disease classification and implemented custom image preprocessing pipelines. Deployed model on AWS Sagemaker with Triton Inference Server achieving 20ms inference time.

Built AI research assistant using GPT-4 and implemented RAG architecture with custom knowledge base. Fine-tuned BioBERT for NER achieving 90% F1-score.

Experimented with models for entity extraction and implemented clustering-based engine. Used XGBoost for classification achieving 85% F1-score.

Developed time series forecasting models for retail sales and leveraged SHAP-values for error analysis. Achieved 12.3% MAPE.

Created automatic spellchecker using TesseractOCR and FastText word embeddings.

Professional Certificate in GCP Professional Machine Learning Engineer from Google Cloud Platform, completed from October to December 2024.

Bachelor's degree in Applied Mathematics from Belarusian State University, completed from September 2020 to June 2024.