Ольга

AI Engineer • Olha Kalynar
Lviv, Ukraine • [відкрити контакти](див. вище в блоці «контактна інформація») • [відкрити контакти](див. вище в блоці «контактна інформація»)
INTERNSHIP
SoftServe Data Science Internship Jul 2025 - Oct 2025
Worked on a GenAI application for personalized travel planning.
● Built modular itinerary parsing system with LLM-first strategy and regex-based fallback
● Implemented production guardrails: budget enforcement, EXIF stripping for image privacy, CloudWatch cost
monitoring
● Built authentication layer using AWS Cognito.
● Developed an interactive Streamlit interface.

TECHNICAL EXPERIENCE
Sleep Quality Prediction Model
● Built regression models using Python, scikit-learn, pandas to predict sleep efficiency with 99.95% accuracy
● Implemented data preprocessing including feature engineering, data cleaning, and categorical encoding
● Compared 11 machine learning algorithms (Linear Regression, Random Forest, XGBoost, Gradient Boosting)
● Optimized hyperparameters using GridSearchCV and cross-validation techniques

CNN Classification
● Built custom CNN architecture using PyTorch for multi-class image classification of anime characters
● Designed convolutional layers with MaxPooling, ReLU activation, and dropout regularization
● Used Kaggle environment for GPU acceleration and distributed computing to optimize training time
● Resolved data leakage issues by identifying and merging duplicate classes that caused overfitting

Multi-Layer Perceptron Implementation
● Built binary classification models using both PyTorch and pure NumPy for raisin dataset classification
● Implemented custom MLP architecture with ReLU activation, sigmoid output, and dropout regularization
● Achieved 87.22% accuracy with PyTorch implementation and 86.11% with pure NumPy neural network

Intelligent Adaptive Learning System for Programming
● Built a 5-agent LLM pipeline (grading → gap analysis → RAG retrieval → lesson & question generation) with
FastAPI and asyncio, cutting evaluation time 3.75× via concurrent grading (asyncio.gather)
● Implemented RAG over a per-user knowledge base with Weaviate vector DB and Jina embeddings
● Validated automated grading against expert scores: Cohen's kappa 0.969, MAE 0.010, retrieval Precision@5
0.70
● Built a secure in-browser code sandbox via AST static analysis (subprocess isolation, timeouts, rate limiting);
AWS Cognito auth, React/TS frontend, 138 tests at 91% coverage
EDUCATION

LVIV POLYTECHNIC NATIONAL UNIVERSITY Lviv, Ukraine
BACHELOR OF COMPUTER SCIENCE - INSTITUTE OF COMPUTER SCIENCE AND INFORMATION TECHNOLOGIES 2022-2026

ADDITIONAL INFORMATION
● Technical Skills: Python, SQL, Google Spreadsheets, LangChain, Docker, LangGraph, Pytest, SQLite, Flask,
AWS (Cognito, S3, Secrets Manager, CloudWatch), Terraform, FastAPI, PostgreSQL, Weaviate, RAG, Redis
● Languages: English - Upper-intermediate, Ukrainian - Native