We've covered what an FDE is and what the job actually looks like day to day. The natural next question: should you want this job — and if so, how do you actually get it? The honest fit test FDE work rewards a specific temperament, and it's worth being honest about whether it's yours before you chase it for the salary alone (and the salary is real: average total comp is now reportedly around $238K, with a range from roughly $205K to $486K, and senior FDEs at frontier labs goi

In the last post, we built up the idea of a Forward Deployed Engineer (FDE) from first principles: an engineer who sits with the customer and ships the code, collapsing the usual gap between "understanding the problem" and "building the solution." That's the theory. What does it actually look like on a Tuesday? The traditional software engineer's day Picture a typical SWE at a product company. Their day is structured around internal development: a standup, some deep-focus tim

If you've browsed tech job boards recently, you've probably noticed a title popping up everywhere: Forward Deployed Engineer. Palantir, OpenAI, Anthropic, Google, Databricks — all hiring for it, some by the dozens. Anthropic even set up a joint venture with Blackstone and Goldman Sachs just to embed these engineers inside financial firms. So what actually is this job, and why does it suddenly matter so much? Let's build it up from scratch. The problem: some software can't jus

When an AI project fails, the post-mortem almost always lands on the model — the wrong architecture, insufficient training data, a tuning choice that didn't generalise. Rarely does the conversation turn upstream, to the input layer where data was collected, cleaned, and structured before it ever reached the model. That omission is expensive. When an AI project fails, the post-mortem almost always lands on the model — the wrong architecture, insufficient training data, a tunin

There is a conversation happening in every enterprise AI post-mortem that nobody wants to be the first to say out loud: the model wasn't the problem. The model did exactly what it was trained to do. It processed what it received, and it produced an output that was only as reliable as what went in. This is not a new concept. Software engineers have lived by "garbage in, garbage out" for decades. But somewhere in the hype cycle of large language models, the enterprise AI indust

Every enterprise AI pilot looks good. This is not a cynical observation — it is a structural reality of how pilots are designed. The data is curated. The scope is narrow. The use case is selected precisely because it is the one where the existing data is cleanest and the model is most likely to perform well. The demo works. The stakeholders are impressed. The project gets approved for production. Then production begins. And something breaks. Over the course of working with en