Lev — Machine Learning Engineer

Work Experience

ML Engineer (Energy and Charging Platforms AI Integration)

Duration: 1 year

Summary:

• The project focuses on integrating advanced AI capabilities into energy and charging platforms to enable accurate forecasting and intelligent optimization
• It includes predictive models for solar generation, EV charging demand, load behavior, electricity pricing, and battery storage management

Responsibilities:

• Developed and maintained time series forecasting models using statistical Prophet and machine learning approaches including XGBoost.
• Performed feature engineering and data preprocessing to improve forecast accuracy and stability.
• Conducted hyperparameter optimization using Optuna to systematically improve model performance.
• Designed and implemented an MLflow-based experiment tracking setup to manage model versions, parameters, and evaluation metrics.
• Prepared, cleaned, and structured data for modeling and analytics.

Technologies: Python, Pandas, NumPy, SciPy, Scikit-learn, XGBoost, PyTorch, MySQL, AWS, Plotly, Docker

ML Engineer (Jamming System for UAVs)

Duration: 1 year

Summary:

• Developed an anti-jamming system for UAVs focused on detecting and classifying radio-frequency jamming on the communication channel
• The solution combines software-defined radio signal processing with deep learning-based classification

Responsibilities:

• Designed and implemented radio signal modulation and processing pipelines using GNU Radio.
• Simulated and generated jammed and non-jammed RF signals for training and evaluation.
• Built and trained CNN-based models to detect and classify jamming patterns in radio channels.
• Performed signal preprocessing and feature extraction to convert RF data into model-ready representations.

Technologies: Python, GNU Radio, CNNs, RF signal processing, FFT, Spectrograms, NumPy, SciPy, Linux, Git

ML Engineer (Protego AI Call Assistant)

Duration: 1 year

Summary:

• Developing a real-time AI phone call assistant designed to handle live voice interactions with users
• The system processes incoming audio streams, generates context-aware responses, and converts them into natural-sounding speech in real time

Responsibilities:

• Designed and implemented a real-time Text-to-Speech (TTS) module for live phone call interactions.
• Explored, evaluated, and benchmarked multiple TTS and speech-related models to select optimal solutions for latency and audio quality.
• Performed model fine-tuning to improve voice naturalness, pronunciation, and response timing.
• Built and containerized services using Docker, ensuring reproducible and scalable deployments.
• Deployed and managed TTS services on RunPod cloud infrastructure.

Technologies: Python, PyTorch, Text-to-Speech (TTS), real-time audio processing, Docker, RunPod, Linux, Git

ML Engineer (Legal Assistant Powered by RAG System)

Duration: 1 year

Summary:

• Legal assistant powered by a Retrieval-Augmented Generation (RAG) system for US constitutional and federal law
• Enabled users to ask complex legal questions and receive accurate, citation-based responses grounded in the US
• Constitution, federal law, and New York State regulations

Responsibilities:

• Parsed and embedded legal documents (U.S. Constitution, federal, and NY state laws) using advanced text embedding models from HuggingFace.
• Built a conversational interface that allows users to naturally query legal texts.
• Implemented agent-based intelligent query reformulation to match user intent with optimal legal passages.
• Ensured answer trustworthiness by grounding outputs in retrieved legal content and providing traceable citations.
• Evaluated multiple embedding and retrieval strategies to boost accuracy and reliability in legal queries.

Technologies: Python, HuggingFace, Langchain, OpenAI, AWS, RAG

Education

• Kyiv School of Economics
  Masters of Computer Science
  
• Lev Bachelor of Programs Engineering