Nurperi Bazlakunova

AI engineer

IRCCS: National institute of oncology in Milan

Built and evaluated a hybrid LLM and symbolic system for oncology trial matching; reported matching metrics and assignments for 110 patients.

• Building a hybrid LLM and neuro-symbolic AI system. • Achieved accuracy 0.87, F1-score 0.7, and ROC-AUC 0.77 in patient-trial matching experiments. • Assigned trials for 110 oncology patients using LLM reasoning and symbolic medical rules. • Worked on evaluation, dataset preparation, and explainable matching logic for sensitive healthcare data.

Data scientist

Bank Kompanion

Top 5 banks in Kyrgyzstan

Built credit-service APIs and customer models; led a card-recommendation project that raised conversion from 3.0% to 3.5%.

• Built FastAPI endpoints for a credit service. • Analyzed customer and loan data from DWH using PostgreSQL/MySQL. • Tracked experiments with MLflow and stored model/artifacts in AWS S3 (artifact store). • Led a card-recommendation model project that increased the conversion rate from 3.0% to 3.5%, a 0.5 percentage-point increase, equivalent to a 16.7% relative uplift. • Built credit-scoring and churn-prediction models using XGBoost and LightGBM on customer and loan data.

AI engineer

Motion Web

Educational platform

Supported model development and advised on architecture choices at an educational platform.

• Supported models development and advised on architecture choices.

Bachelor, Artificial intelligence

University of Milan (UNIMI)