Jeremiasz Wołosiuk

Led a 5-member team of software engineers and researchers in developing superhuman-level AI for imperfect-information games.

Supervised and actively contributed to implementing SOTA imperfect-information game AI (Deepstack, ReBeL, Modicum, Pluribus) using Neural Networks and depth-limited evaluation techniques.

Conducted advanced research to exceed SOTA performance, focusing on novel improvements and opponent exploitation strategies.

Managed strategic planning, roadmap development, and stakeholder communication, ensuring swift development and optimal research trade-offs.

Directed a 5-member engineering team in developing the core game-solving Engine, ensuring algorithmic soundness based on advanced papers and game theory.

Recruited and onboarded new team members while actively contributing as a software developer, data scientist, and ML engineer.

Implemented significant speed optimizations across PyTorch code and Neural Network training through profiling tools, enhancing system performance.

Pioneered the first commercial Poker solver utilizing neural networks, with algorithm improvements recognized by a top AI conference.

Developed a robust Poker solver from scratch using CounterFactual Regret algorithms, implemented in Python (PyTorch) with performance-heavy parts in C++.

Conducted intensive State-of-the-Art (SOTA) research in game AI and collaborated effectively within a 3-person research team.

Engineered robust and scalable solutions for complex game environments, demonstrating expertise in both theoretical AI and practical software development.

Collaborated to automate marketing campaigns on Google Ads and Facebook Ads, involving web scraping, ETL design (Docker, Celery), and NLP data analysis.

Implemented efficient ETL processes, streamlining data analysis workflows and significantly improving marketing campaign performance.

Contributed to the development of innovative features for the marketing application, enhancing user experience and campaign effectiveness.

Designed and executed experiments to predict players' fear/boredom emotions in horror games, providing data-driven insights for design.

Analyzed extensive in-game behavior data and applied machine learning models to extract actionable insights for product improvement.

Contributed directly to enhancing player experience and optimizing game design for QED Software through data-driven recommendations.

Developed secure and robust database applications in .NET, ensuring high data integrity and security for client systems.

Collaborated with team members to efficiently troubleshoot and resolve critical customer database issues, minimizing downtime.

Implemented software solutions that enhanced overall system efficiency and improved user experience.