Andrii Balashov

I’m a Senior AI/ML Engineer and Reinforcement Learning Researcher based in Ann Arbor, MI. I’ve spent more than 10 years building large-scale ML systems, and these days I’m especially interested in Deep Learning, RL, and turning research ideas into tools that work in the real world.

I conduct my research in the AI Lab at the Ukrainian State University of Science and Technology. I have an M.S. in Computer Science, review research papers for Elsevier, and lead the WordPress AI team.

Keywords: Large Language Models, Reinforcement Learning, RLHF, LLM Security, Deep Learning, Graph Neural Networks, MLOps

Credentials: M.S. Computer Science · AWS Certified ML – Specialty · Azure AI Engineer · Elsevier peer reviewer · WordPress Core AI lead

Publications

TriRoute: Unified Learned Routing for Joint Adaptive Attention, Experts, and KV-Cache Allocation

TriRoute: Unified Learned Routing for Joint Adaptive Attention, Experts, and KV-Cache Allocation

arXiv:2607.06601, 2026

A Spatio-Temporal Graph Attention Network for Real-Time Damage Localization in Prestressed Concrete Bridge Structures under Operational and Seismic Loads

A Spatio-Temporal Graph Attention Network for Real-Time Damage Localization in Prestressed Concrete Bridge Structures under Operational and Seismic Loads

Journal of Physics: Conference Series 3251 (1), 012003, 2026

Probabilistic Deep Learning for Real-Time Seismic Damage Assessment in Reinforced Concrete Structures

Probabilistic Deep Learning for Real-Time Seismic Damage Assessment in Reinforced Concrete Structures

Journal of Physics: Conference Series 3251 (1), 012002, 2026

Adaptive Traffic Signal Optimization with Thermal Sensors and Reinforcement Learning

Adaptive Traffic Signal Optimization with Thermal Sensors and Reinforcement Learning

Results in Engineering, 2025

Reinforcement Learning Fine-Tunes a Sparse Subnetwork in Large Language Models

Reinforcement Learning Fine-Tunes a Sparse Subnetwork in Large Language Models

arXiv:2507.17107, 2025

Multi-Stage Prompt Inference Attacks on Enterprise LLM Systems

Multi-Stage Prompt Inference Attacks on Enterprise LLM Systems

arXiv:2507.15613, 2025

Projects

ReturnFast

ReturnFast

Native macOS utility that keeps one browser-style Back/Forward trail across every app.

WebsiteHow it worksSupported appsProduct Hunt

WordPress Core (AI Team)

WordPress Core (AI Team)

AI team lead and core contributor shaping AI features for a platform powering ~40% of the web.

Make.WordPress.orgGitHubWordPress 6.5 "Regina"

Social Warfare

React/PHP publishing plugin with 1M+ downloads, serving ML-driven content ranking at scale.

WebsiteGitHub (Core)GitHub (Pro)WordPress.org

Denial Shield

Denial Shield

Medical claims platform cutting denials 75% with transformer document understanding and calibrated risk scoring.