AI Application Development
AI product prototypes, application architecture, evaluation feedback, and real user scenarios that turn AI into usable software capability.
About Roark Li
A decade of engineering toward software worth keeping.
Roark Li is a senior software engineer and long-form writer. I entered software through cross-disciplinary self-study, built product platforms and developer infrastructure at Alibaba, and now focus on AI applications, independent products, and technical writing.
Career Throughline
From software engineering to product systems and AI applications.
My story is not about chasing labels. I keep turning ideas into usable software: first products and tools, then developer platforms and infrastructure, and now AI applications, independent products, and technical content. The object changes, but the standard remains: clear goals, reliable engineering, thoughtful architecture, and human judgment.
- 01
Software Engineering
From self-study to professional practice
- 02
Developer Productivity
Tools that improve how teams build
- 03
Platform Engineering
Systems for delivery at scale
- 04
AI Applications
Turning model capability into products
In Progress
Current Focus
Software Thinking
Ongoing thinking on AI coding, software quality, developer capability, and how architectural judgment changes the way software is made.
Products and Technical Content
Public projects, technical essays, and tutorials that make experience readable, testable, and open to continued discussion.
Selected Evidence
Some public proof
Published by Alibaba Group
Skill Up
An open-source CLI framework published by Alibaba Group for evaluating AI Agent Skills. While at Alibaba, I was responsible for its overall design and core evaluation capabilities and contributed to maintenance.
View project →Developer Infrastructure
Persistent Build Cache
A frontend build cache service that accelerated builds about 130,000 times across 2,000+ projects between 2024 and 2025.
Product Platform
PopLayer
A Taobao no-code marketing popup platform upgrade that covered nearly 90% of marketing popup scenarios and supported large-scale e-commerce campaigns.
Read retrospective →Principles
I believe AI changes how software is built, not judgment itself.
- 01AI can lower the cost of generating code, but it does not automatically create good products.
- 02Useful AI applications need product judgment, engineering quality, feedback loops, and long-term ownership.
- 03Technical content and projects matter because they reveal how someone thinks, how they work, and whether they are worth trusting.
Keep reading, or stay in touch
Subscribe if you also care about AI applications, software quality, product judgment, and independent creation. Specific conversations about writing, podcasts, or projects are welcome by email.