top of page
What is Google Mainframe Modernization?
Overview
Mainframes are mission-critical supercomputers powering 80% of the world’s data and supporting 92% of top global banks. To bridge this gap, we built Google Cloud Mainframe Modernization—an AI-powered, end-to-end solution that structures legacy transformation into a predictable 4-step workflow: Assess, Modernize, Test & De-risk, and Data Migrate.
Due to NDA, specific process are confidential. Let's connect to discuss my process, testing methodology, and strategic impact!
Context
What role do I play and how do I collaborate with others?
Role & Team
Sole Product Design Lead for Google Cloud’s Mainframe Modernization portfolio, owning end-to-end UX for core AI products (MAT/Dual Run) and scaling them from MVP to enterprise maturity. Collaborated globally (US/IL/IN) to co-author PRDs and pioneered vibe coding for rapid engineering alignment on complex data.
Problem
How can customers trust opaque AI when a single error costs millions?
From Black Boxes
Step 1
into crystal boxes
Structured and delivered an incremental status engine that visualizes the step-by-step progression of the AI's background analysis, exposing real-time execution status, current focus, and instant error diagnostics.
Built AI and human trust by replacing waiting anxiety with rich, incremental updates and reassuring telemetry.
Context
User-in-Loop
Step 2
AI results
Anytime context injection for higher precision. Designed an intuitive "Edit by adding context" workflow, allowing users to inject custom business rules or supporting files at any stage to steer and refine AI results on demand.
Built AI and human trust by enabling real-time, bi-directional steering.
Context
Co-Steering AI
Step 3
with active context
Designed an intuitive pipeline translating cryptic legacy code into plain-English rules. By introducing seamless edit and validate workflows, the UX empowers non-technical stakeholders to audit and control AI outputs.
Built AI and human trust by transforming users from passive observers into active system calibrators who govern AI precision in real time.
User-trained
Step 4
language models
Refining existing and training new languages. Designed a flexible dual-pathway interface that empowers users to customize standard language profiles or bootstrap new language rules, bringing unmatched adaptability to code modernization.Built AI and human trust by democratizing model customization.
Agentic Chat
Step 5
as an assistant
Generative data visualization on demand. Upgraded the existing assistant with agentic capabilities, enabling users to instantly generate, customize, and pivot visual data charts purely through natural language queries.
Built AI and human trust by providing proactive, context-aware assistance.
bottom of page