I'm Aliang, a software engineer and a graduate of South China University of Technology (SCUT). I build intelligent, data-driven systems at the intersection of recommendation systems, AI agents, AIGC, and software engineering.

I care about the full path from an idea to a dependable product: problem definition, experimentation, model and workflow design, evaluation, observability, and production reliability.

• Recommendation systems — retrieval, ranking, personalization, user-interest modeling, and online/offline architecture
• Agentic systems — tool use, planning, memory, context engineering, durable execution, and human oversight
• AIGC — grounded and controllable generation across text, image, audio, and video
• AI engineering — evaluation, structured outputs, guardrails, observability, latency, cost, and maintainability

• adgen_agent — an experiment in agent-assisted advertising content generation
• recommender_system — recommendation-system experiments covering offline, nearline, and online components
• search_sim_img — a Python project for similarity-based image retrieval
• quant-system — a Python-based exploration of quantitative systems

These repositories reflect ongoing experiments and learning. I prefer shipping small, inspectable systems and improving them through real feedback rather than treating a prototype as a finished product.

• Agent harness engineering — the runtime layer connecting models with tools, environments, memory, permissions, observability, and verification
• Hermes Agent and extensible personal-agent ecosystems
• MCP, A2A, and Agent Skills for reusable tools, interoperability, and specialized workflows
• Self-evolving agents, multi-agent orchestration, long-running tasks, checkpoints, retries, and trace-based evaluation
• Coding agents that can understand repositories, implement changes, run tests, and provide evidence-backed results

• Text-to-video and image-to-video generation
• Camera, motion, timing, composition, and style control
• Temporal consistency and character continuity across multi-shot sequences
• Multimodal storytelling across scripts, images, speech, music, and sound effects
• Agent-assisted workflows for research, storyboarding, asset generation, editing, and quality review
• Evaluation of prompt alignment, motion quality, visual fidelity, continuity, safety, latency, and cost

• Modern retrieval and ranking architectures
• Evaluation and iterative improvement of personalization quality
• RAG, knowledge-grounded generation, memory, and context-management strategies
• The intersection of LLMs, agents, and recommendation systems

• Define the problem and measurable success criteria before choosing a solution
• Use experiments and evidence to guide technical decisions
• Balance quality with latency, cost, complexity, and maintainability
• Prefer simple, explainable designs with clear execution boundaries
• Build observability, verification, and human control into AI workflows
• Document lessons and contribute useful ideas back to open source

An automatically updated snapshot of the languages I've worked with over the past seven days, powered by WakaTime. It reflects hands-on coding activity rather than the full scope of research, design, and technical exploration behind the work.

From: 19 July 2026 - To: 26 July 2026

Total Time: 0 secs

No activity tracked

I'm open to thoughtful conversations and collaboration around recommendation systems, agents, AIGC, and practical AI engineering.

• Email: linxingliang@163.com
• GitHub: @Aliang-CN