Agenda
Dual-Process AI Agents in Production: Automating QA and Enhancing PGO at Capcom
Kazuki Abe (Technical Director and Head of AI Solutions and Platform at Capcom, CAPCOM CO., LTD.)
Kota Fukasawa (Head of Game Technology & Platform Engineering at Capcom, CAPCOM CO., LTD.)
Pass Type:
Academic Pass, Festival Pass, Game Changer Pass, Indie Pass, Start-Up Pass -
Get your pass now!
Hot Topics:
AI
Track:
Programming, Production & Team Leadership
Audience Level:
Advanced
Format:
Lecture
Vault Recording: TBD
Audience Level: Advanced
To overcome massive testing bottlenecks, a cross-functional scrum uniting Capcom's game developers and shared technology engineers built a production-proven AI framework. This proactive collaboration automates 45,000 monthly hours of game QA and enhances Profile-Guided Optimization (PGO) by 50% across major titles like Monster Hunter Stories 3 and Mega Man: Dual Override. Using a dual-process neuro-symbolic architecture, autonomous AI agents play games using only raw video and audio inputs to detect performance bottlenecks, visual bugs, asset redundancies, and text errors. Gameplay profiles feed directly into native PGO compilers to maximize efficiency. By bridging AI engineering with application integration, this agile team delivers automated, actionable fixes straight to developers via Model Context Protocol (MCP). Attendees will learn how cross-functional collaboration and AI integration can revolutionize modern game development.
Takeaway
Attendees learn to build dual-process neuro-symbolic AI agents operating solely on raw audio-video inputs. They discover how autonomous execution logs boost PGO compiler efficiency by 50% and how to deliver actionable fix proposals for performance and visual artifacts directly into developer workflows via Model Context Protocol.
Intended Audience
This session targets AI engineers, engine programmers, tools developers, and QA leads seeking advanced automation. Attendees benefit by learning cross-title AI architectures and automated profiling strategies. Prerequisite knowledge includes basic familiarity with game engine pipelines, machine learning fundamentals, and code optimization concepts.