Data scientist

OMON

About the Project

Building a Visual Inertial Navigation System (VINS) for copter drones that need to fly without GPS. The system uses camera images combined with IMU measurements to calculate position autonomously. In simulation, it achieves 4m error on 1km flights and 250m error on 20km flights. In real-world testing, the copter completes 5km missions with under 500m maximum error. The project also includes a map module with keypoint precomputation, georeferencing, and visual feature matching.

• Defense
• UAVs/drones
• Navigation

Responsibilities

Worked on algorithm tuning for the OpenVINS-based navigation system, adapting parameters to improve accuracy across different flight conditions. Handled system development and debugging, including integration with ROS and ArduPilot. Performed extensive log analysis from both simulated and real flights to identify error patterns and guide improvements. Conducted testing within the AirSim simulation environment, validating navigation accuracy before real-world flight trials. Worked with C++ and Lua alongside Python for the core navigation components.

MURKA GAMES LIMITED

About the Project

Murka Games is an international studio producing social casino and casual games for over a decade. The data science work focused on analyzing player behavior to find segmentation opportunities, forecast in-game actions, and optimize monetization strategies. This covered the full analytics cycle: understanding why players spend (or stop spending), predicting churn before it happens, and timing in-game offers for maximum conversion. The work directly influenced game design decisions and marketing spend allocation.

• GameDev
• AdTech
• Analytics

Responsibilities

Spent 38 months running the full ML lifecycle for player analytics. Built machine learning models for behavior forecasting, customer segmentation, and marketing targeting, handling everything from data preprocessing and exploratory analysis through model development, production deployment, and ongoing monitoring. Analyzed A/B test results to validate game mechanics changes. Developed supporting tools including decision support systems, automated HTML reports, and trigger systems for in-game mechanics timing. Presented findings and model results to management regularly.

ARX FAIRFAX

About the Project

Fairfax is a major property and casualty insurance and reinsurance company. The project focused on processing and analyzing actuarial data, with specific responsibility for calculating and forming technical reserves. This also included building the division's BI reporting infrastructure from scratch, and automating information extraction from unstructured insurance documents like policy texts and claims narratives using Python NLP techniques.

• Insurance
• Fintech
• Analytics

Responsibilities

Created a comprehensive BI reporting system for the division using Power BI. Implemented Python-based automated extraction of structured data from unstructured policy and claims text, improving both accuracy and processing speed. Handled routine actuarial data processing and analysis for reserve calculations. Managed software development projects across the full lifecycle, from requirements gathering through testing. The automation work was part of a larger efficiency initiative across the division.