System framing
Turning unclear needs into modules, data ownership, failure paths, and decisions that a team can actually build from.
RequirementsArchitectureReview
Hello, I'm
Xiaonan Li
AI Systems & Full-Stack Engineer
I design the system, then use AI to move faster.
Software engineering student focused on turning unclear needs into service boundaries, AI worker flows, APIs, and product behavior that can survive review, testing, and real users.
Current signal
Interested in teams where AI capability needs architecture, ownership, and real software discipline.
Explore the systemxiaonan.dev
const profile = {
name: "Xiaon01 / Profile
From writing code faster to designing better systems.
I use AI heavily, but the part I care about most is deciding what should exist: how to split modules, where state belongs, how workers coordinate, which failures matter, and how product needs become reliable software mechanisms.
I communicate in Chinese and English, and I am actively building Japanese ability for broader collaboration and technical reading.
AI worker design
Designing pull-based task flows where AI services claim work, report state, and stay separated from business APIs.
WorkerAsyncBoundary
Full-stack execution
Using backend, frontend, Android, and deployment work to turn the design into something usable after the demo.
Spring BootNext.jsAndroid
Languages
Chinese, fluent English for technical collaboration, and Japanese as an active learning focus.
中文English日本語
02 / Experience
Internship work read as system responsibility.
Each role is presented by scope and impact rather than a collapsing resume card, because the important part is what had to be understood, separated, debugged, and made reliable.
Meituan Platform
Software Development Engineer Intern
Worked on reliability, status reporting, debugging, dashboards, and build workflow optimization in core business systems.- 01
Designed and optimized system status reporting and retry mechanisms under unstable network conditions.
- 02
Used AI-assisted workflows for complex issue analysis and system behavior investigation.
- 03
Used JADX to reverse engineer and debug obfuscated code, improving reporting format reliability.
- 04
Built Spark and SQL dashboards for anomaly detection and trend analysis.
- 05
Analyzed Gradle task DAG bottlenecks and optimized build workflows while maintaining stability.
03 / Education
Academic foundation across software engineering and applied ML.
Coursework and research practice connect software engineering fundamentals with applied machine learning and team-based product delivery.
Swinburne University of Technology
Bachelor of Engineering (Honours)Software Engineering
Software engineering, applied machine learning, product thinking, and team delivery come together here through coursework and project practice.
Software EngineeringApplied MLTeam Delivery
Shandong University of Science and Technology
Bachelor of EngineeringSoftware Engineering
Built the computer science foundation across algorithms, operating systems, databases, software engineering, and thesis research on federated learning coordination.
CS FundamentalsFederated LearningDatabases
Scroll horizontally between schools
04 / Projects
Projects where architecture is the proof.
Selected work where the value is not only generated code, but the system decisions behind AI workers, retrieval, mobile constraints, and model evaluation.
Choose a project here; the switcher stays pinned while you inspect the blueprint.
AI-driven Enterprise SaaS Platform
Enterprise AI SaaS with worker-driven analysis flows for knowledge extraction and structured decision support.
Problem
Customer-facing AI analysis cannot stay as a one-shot request when documents, evidence, retries, and long-running tasks are involved.
System decision
I proposed a pull-based AI worker model: backend stores tasks, AI workers claim work, report state, and keep business APIs decoupled.
Proof
The architecture separates Spring Boot APIs, Python AI workers, persistence, callback state, and deployment paths so changes can be reviewed and tested by contract.
Stack
AI Worker, RAG, Spring Boot, Python services, PostgreSQL, AWS.
01Led architecture design, task decomposition, and full-stack development for an enterprise AI SaaS platform.
02Proposed a pull-based AI worker model where AI services claim backend tasks, report state, and avoid tight request coupling.
03Designed a RAG pipeline with parsing, chunking, embeddings, semantic retrieval, and dynamic context injection.
04Built an OpenAI-compatible LLM abstraction layer with model switching, streaming, and schema-constrained JSON generation.
Project01 / 03AI-driven Enterprise SaaS Platform
05 / Stack
Stack, but written as applied capability.
The tools below are grouped by the system decisions they support: orchestration, service boundaries, product interfaces, testing, and deployment.
01
AI & Machine Learning
Used for retrieval, model integration, visual inference, and training workflows.
PyTorchTensorFlowTensorFlow LiteRAGFAISSSentenceTransformers
02
Backend & Data
Used for service architecture, APIs, persistence, data processing, and deployment.
Spring BootJavaMyBatisMicroservicesSparkSQLAWS
03
Frontend & Mobile
Used for product interfaces, Android SDK work, testing, and portfolio delivery.
Next.jsTypeScriptTailwind CSSAndroidKotlinJUnit
06 / Contact
Let's build something that survives contact with reality.
Open to teams that need AI ideas translated into architecture, implementation plans, reviewable code, and software that keeps working after the first demo.
Daily build note
Prototype fast. Land it carefully.
The demo matters, but the real signal is whether people can use it after the room gets quiet.