Forward Deployed Engineering
Taking AI into production:
from demo to a system you can accept
We are a forward deployed engineering (FDE) firm: we embed engineers with your team to wire AI into the systems, permissions and workflows you already run. Most enterprise AI pilots stop at the demo — the model is rarely the problem.
- We build AI products ourselves
- kduanju.com — our own AI short-drama production SaaS, running on 10+ vendors and 40+ models
- Coverage
- On-site in Beijing; remote across China and worldwide
- Legal entity
- Zhiyuan Qidian (Beijing) Intelligent Technology Co., Ltd.京ICP备2026041781号-3
We do one thing: close that gap
Today
Pilots everywhere, nothing in production
- The demo looks great until it meets real data
- Business cannot specify it; IT does not understand it
- Model choice keeps being reopened, nobody owns the outcome
- Front-line staff were never consulted, so the system sits unused
After
One system running in production, with users
- Wired into your existing systems, permissions and workflows
- Agreed acceptance criteria and a measured result
- Front-line staff shaped it rather than receiving it
- Left behind as reusable components — the next case goes faster
How we work together
01 / Assess
Decide whether it is worth doing
One sixty-minute assessment covering the use case, the data and the organisation. If the honest answer is that you should not build this, we will say so.
02 / Pilot
Prove it small, on real data
A fixed-scope pilot running against your real data with your real users, accepted against criteria agreed up front.
03 / Compound
The knowledge stays
Components, connectors and workflow templates remain available to you — and make the next use case faster to deliver.
Three questions you will ask before handing us the work
We answer them with commitments that can go into a contract, not adjectives.
Named people, in writing
Will the people who do the work be the people I am talking to?
Yes. We name the team in the proposal and fix it at quote stage. If someone has to be replaced mid-project we tell you in advance and explain why; if the change undermines the project, you can decide at the end of that phase whether to continue.
Data boundary in the contract
Will our data be used somewhere else?
No. Client data, business rules and system detail do not enter our reusable assets — what is reused is the general method and components with that information removed. That is stated in the contract and enforced in how we work. Where data is sensitive, we can work only inside a network you control.
Limits stated up front
What can you actually do, and what can you not?
We do not sell headcount by the day. We do not take work that is really a request for cheap labour. We do not promise to replace your core systems, and we do not promise unverifiable numbers like "30% efficiency gain". When a use case is not worth doing, we write that down — which costs you less than a project that should never have started.
Common questions
What does an AI deployment firm actually do?
We connect general AI capability to the specific systems, data, permissions and workflows of one organisation, and get it running in production. Companies are rarely short of ideas or of models — what blocks them is the middle stretch: whether real data can be connected, whether the process can absorb it, who is accountable when it goes wrong, and who maintains it afterwards. That stretch is our work.
Should this not be quick, with modern AI?
A demo is quick — weeks, sometimes days. The hard part is the distance from prototype to production: real data arrives in several formats with large gaps, the target system may not expose an interface, permissions need approval at several levels, and front-line staff may not want to change how they work. More than 60% of enterprise AI attempts stall in pilot, and the blockers sit in those places rather than in model capability. So we will not promise you a fast launch — that promise tends to resurface as a dispute at acceptance.
How is this different from ordinary outsourcing?
It comes down to what remains when the project ends. If all you are left with is a system that runs, that is outsourcing. If the vendor takes knowledge away but cannot reuse it, that is project work. If the domain knowledge — stripped of your data — becomes reusable components and templates, that is forward deployed engineering. The test is concrete: if the first customer takes ten person-months and the tenth customer in the same industry still takes ten, the model has not worked.
Do we actually need a forward deployed engineer?
Often you do not. If your requirement is standardised and an off-the-shelf product covers it, buying that product is cheaper. Forward deployed engineering fits requirements that are not standard, that touch several existing systems, and where someone has to help you decide what to build first. We start with an assessment, and if the answer is that you should not do this, we will say so.
Can you work on-site?
On-site in Beijing. Elsewhere in China and internationally we work remotely.
How do you charge?
Not by the day rate. We work through a fixed-scope pilot, staged acceptance, or an outcome-based retainer, depending on how well-defined the requirement is. See the engagements page for detail.
Request an FDE assessment
We start with 60 minutes to understand your situation and judge whether this is worth doing, and whether it is worth doing with us. If the answer is no, we will say so.
Fill this in, then press the button below: the email is copied to your clipboard. Paste it into your mailbox and send it to the address above. Nothing is uploaded to our servers.
Start with the project that is stuck
No deck needed, and you do not have to know the solution yet. Describe where it is stuck; we will tell you whether we can help and whether it is worth doing.