Do ordinary people have a chance to enter the FDE? What abilities do I need to supplement? How to prepare for the interview? And-whether this position is worth investing in.
Give the conclusion first:
The entry threshold for FDE is lower than most people think; but it is by no means a simple job with more money and less work.
Many people can't enter this field. It's not that they are not capable enough, but that they set themselves on the standards of senior engineers from the beginning.
In fact, FDE is more like a comprehensive position:
You need to understand some technology, but you don't need to be an algorithm expert;
You need to understand the business, but not necessarily from a traditional consulting background;
You need to be able to communicate with customers, but not just sales.
What really opens the gap is whether you can turn a vague problem into a workable, measurable, and replicable solution.
This is also why some people are very skilled but are not suitable for FDE; while some people with ordinary engineering skills but are good at solving real problems grow up quickly.
1. Three paths to enter the FDE
If sorted by recommendation level:
Give priority to internships/school recruitment entrances, and at the same time, no matter which path you take, you must do real projects.
Because the FDE ultimately examines not "what have you learned", but:
Have you ever really solved someone else's problems?
Path A: Internships and new entrance-the most easily underestimated opportunities
Many people think that FDE is a position that can only be filled after several years of experience, but in fact, many companies have opened up internships and school recruitment.
Recruitment requirements are often not as high as expected.
The core is to look at three things:
The first is whether you have actually used AI.
Instead of simply asking a few questions, it's:
- What tasks have I accomplished with AI?
- Have you changed the original workflow?
- Have practical problems been solved?
第二，是否具备基本工程能力。
不要求你成为高级开发，但至少需要：
- 能写 Python 脚本；
- 能调用 API；
- 能搭建简单应用；
- 遇到问题知道如何自己排查。
第三，你有没有主动做过项目。
课程作业、个人项目、黑客马拉松，甚至帮助别人解决实际问题，都可以成为证明。
因为招聘方真正想知道的是：
你是不是一个会主动把事情做出来的人。
路径 C：自己完成一次真实交付——这是主线
无论你是否走实习、校招或者政策通道，这条路都绕不开。
因为 FDE 面试最核心的问题永远是：
“你过去解决过什么问题？”
Instead of:
“你看过哪些课程？”
学历、证书、培训经历，都无法替代一次真实交付。
一个合格的 FDE 项目，至少应该满足：
- 使用真实数据；
- AI 进入真实业务流程，而不是停留在聊天；
- 有明确结果；
- 能部署运行；
- 有真实用户使用；
- 能衡量效果。
其中最重要的是最后两点。
很多人可以做 Demo，但无法完成交付。
真正区别在于：
Demo 证明你会做东西；
交付证明你能解决问题。
第一个真实客户，比第一个技术项目更重要
很多人学习 AI 的路径是：
学模型 → 学框架 → 做 Demo → 找工作。
但 FDE 更接近：
发现问题 → 设计方案 → 快速验证 → 交付结果。
所以你需要尽早面对真实用户。
第一个客户不一定需要收费。
它的价值不是收入，而是让你经历真实环境：
客户不会准确描述需求；
数据通常是不完整的；
The process must be more complex than imagined;
Things will definitely change halfway through.
These experiences cannot be gained by reading tutorials.
The way to find customers is actually very simple:
- Your past industry connections;
- Small businesses without technical teams;
- Individuals or teams who often complain about duplication of work.
The first cooperation can be cheap or even free, but it must be clear:
Delivery of what?
How to accept it?
How to judge the effect?
When you are done, record the process.
Because this copy is the most valuable resume material for you in the future.
The biggest difference between FDE and ordinary implementation is not how much technology is used.
Instead:
Can you identify the real problem and build a method that you can replicate later?
This is also the core ability of the FDE.