Hi, I’m GuanLin π#
AI Architect, focused on Enterprise AI system design in the tech industry. If you’re wondering “we finished the AI PoC β now what?”, you might find something useful here.
My core work is one very concrete thing: transforming enterprise AI PoCs into infrastructure that actually runs in production. Not demos, not reports β but systems that can sustain daily operations, are maintainable, and can evolve over time.
Along this path to production, the breaking points are rarely just technical:
- Engineering vs. Reality: Clean test data becomes messy noise in real environments; security and compliance issues you never considered in local development surface when moving to the cloud or a Hybrid Cloud setup.
- Organization vs. Expectation: Frontline employees resist change out of fear of being replaced, so even after automating workflows, the way of working stays the same; decision-makers tend to mythologize AI, and when expectations collide with technical reality, projects lose support midway through.
Technology can always iterate, but if the friction between engineering and organization isn’t addressed, even the most perfect architecture won’t make it to production.
What I Focus On#
I follow the full journey of Enterprise AI from technical research to commercial deployment. Here are the areas I actively practice and track:
- Enterprise RAG Systems: How to make LLMs accurately understand heterogeneous internal documents β complex regulations, financial reports, and multi-modal scanned files
- Agentic Systems & Inference Optimization: From multi-agent collaborative architectures to inference performance optimization in production, making models deliver reliable results in complex business scenarios
- On-Premise & Cloud Infrastructure: The last-mile architectural considerations for AI deployment β security compliance, hybrid cloud architecture, and operational cost control
- AI Product Strategy & Expectation Management: Building great technology that nobody uses is the most common failure mode in enterprise. Managing stakeholder expectations is an invisible but critical part of architecture design
- Modern Data Engineering: Data is the foundation of AI. Without clean, structured Data Pipelines, even the most powerful models are built on sand
The Story Behind This Blog#
GuanLin’s Latent Space is named after the machine learning concept of “Latent Space” β a high-dimensional semantic space that carries a model’s understanding of the world.
I want this blog to be my own latent space: leaving a trace of every learning journey, every implementation pitfall, every piece of content that changed how I think.
Not because I already know everything, but because making the learning process public has intrinsic value β it’s a record for my future self, and perhaps a useful reference point for whoever happens to find it.
You’ll find research notes, architecture thinking, implementation logs, and occasionally broader observations on AI deployment in practice.
Get in Touch#
If anything here sparks an idea, or you’re dealing with similar Enterprise AI challenges, feel free to reach out on LinkedIn.