About Yunshang AI

Who I am

I come from an engineering background: I have been writing code for over a decade, and I have also built hardware.

In 2024 I started putting large language models into real business workflows, and one thing became clear fast: model capability is not the bottleneck — accuracy is. With the same model, a well-built knowledge base versus a poor one can differ by more than 30 points in accuracy.

So I stopped selling « AI solutions » and now do one narrower thing: make your AI answer accurately, and produce evidence of exactly how much better it got.

What I believe

1. I do not promise what cannot be delivered. I will not say 100% accurate, and I will not say « deploy once, works forever ». Knowledge changes and models degrade — that is why this work has to be ongoing.

2. Metrics come before the plan. Before discussing how, we define what « good » means. Projects where no metric can be defined, I turn down.

3. Private whenever possible. Your expertise and your data are your assets; they should not become fuel for training someone else’s model.

How I work

• Health check first, then talk cooperation: quoting without a check is not serious
• Every stage has an acceptance-ready deliverable
• Monthly metric reports, including the parts that got worse

Credentials and compliance

Put here: business licence, ICP filing number, software copyrights, partner organisations. Add whatever you actually have; do not invent anything. Empty is better than fake.

How to reach me

WeChat / Phone: (to be filled)
Email: (to be filled)
Working hours: weekdays 9:00–18:00, messages outside hours answered the next day