Skip to main content
JinzaiLab Ching Hsu's AI automation studio

Hello, I’m Ching.

I turn time-consuming workflows into AI tools your team can actually use.

I’m an AI engineer and researcher based in Taiwan. I work on document processing, research automation, media workflows, and API integrations for teams around the world.

Tell me about your project See selected work

Before anything else

AI can build it. Should you still hire someone?

  • AI often trades the money a project used to cost for the time you put in yourself.
  • Outsourcing can make sense when your plate is full, you want to learn how someone else approaches the work, or you need experience you do not yet have.
  • For a disposable one-off, or when learning is the whole point, doing it yourself is usually the better call.
Read the full argument →

How I can help

We start with your existing process: where the data comes from, who uses the output, and where human review still matters. The technology comes after that.

Workflow automation and internal tools

I turn repetitive work across spreadsheets, email, cloud storage, and manual steps into maintainable workflows and small internal tools.

Typical work: API integrations, n8n workflows, batch processing, internal dashboards, and automated reports.

More about this service →

Document and research automation

I build practical systems for parsing, checking, evaluating, and organizing PDFs, presentations, documents, and research material.

Typical work: document parsing, OCR, evaluation frameworks, research QA, and structured reporting.

More about this service →

AI-assisted media workflows

I connect the repetitive steps around footage intake, transcription, editing, captions, metadata, generation, and delivery.

Typical work: automated editing, subtitle pipelines, video analysis, media APIs, and batch generation.

More about this service →

Selected work

Public summaries of real projects, with client data and sensitive implementation details removed.

All work →

Media workflow

AI video editing and delivery tool

A desktop tool combining speech recognition, LLM-assisted clip analysis, FFmpeg editing, metadata generation, and FTPS delivery.

Reduced editing time by about 80% based on existing production records.

Document intelligence

AI pitch deck evaluation system

A structured review workflow for PDF and PPTX decks, covering team, technology, R&D, business, and presentation quality.

Turned an internal notebook workflow into a product-ready evaluation system.

Workflow automation

Company data validation workflow

An n8n workflow connecting CSV input, structured LLM validation, exception handling, retries, and HTML email reports.

Nineteen nodes provide a traceable and retryable end-to-end process.

A straightforward way to work together

  1. Understand the current process

    We map the workflow, data, constraints, and the outcome you need.

  2. Choose a sensible first step

    Depending on risk and budget, this may be a focused PoC, one integration, or a complete tool.

  3. Deliver and validate in stages

    Each stage is tested with real data and documented for review and handover.

Notes from the work

Practical decisions, limitations, and lessons from AI projects.

All articles →

Have a workflow you would like to improve?

Send me a short note about your current process and the outcome you need. I will help you identify a practical starting point.

Tell me about your project