Applied AI / Forward Deployed Engineer · The Role: Applied AI & Forward Deployed Engineer · Lesson 1 of 5
What an Applied AI / Forward Deployed Engineer is
An Applied AI Engineer (AAE) builds products and features on top of foundation models. A Forward Deployed Engineer (FDE) does the same thing but embedded with a customer — flying to their office (or Zooming into their mess), understanding their workflow, and shipping a working AI solution against real, dirty data on a deadline. The term was coined at Palantir; OpenAI, Anthropic, and a wave of AI startups now hire the role heavily.
The unifying idea: you are the person who turns a general-purpose model into something that solves a specific person's problem in production. You are not training frontier models. You are the bridge between the model's raw capability and a customer's outcome.
The three things every version of this role blends
- Real software engineering — you ship code: APIs, data pipelines, evals, glue. Python is the lingua franca; TypeScript/React for anything customer-facing.
- Applied LLM craft — prompting, RAG, agents, evaluation, cost/latency tuning. You know what models can and can't do and design around the limits.
- Customer / product instinct — you scope ambiguous problems, talk to non-engineers, demo, and iterate. FDE weights this heaviest; pure AAE weights engineering heaviest.
The one-linerResearch engineers make the model smarter. You make the model useful. Interviewers are testing whether you can take a fuzzy business problem and a fallible model and reliably close the gap between them.
Where the role came from — and why it exploded
- Palantir invented it: Deltas (embedded engineers who ship against one customer's mess) paired with Echoes (domain strategists who find the problem worth solving). Palantir's own framing: product engineering = 'one capability, many customers'; FDE = 'one customer, many capabilities.'
- The best one-liner in circulation (Bob McGrew, ex-Palantir/OpenAI): FDEs 'build the rough gravel road' where the product needs to go; product engineers pave it into a highway for the next ten customers. FDE is product discovery, not consulting.
- Why every AI company now hires it: the great majority of enterprise AI pilots fail on integration, not model quality — someone has to close the last mile inside the customer's messy reality. That's you.
Great 'Why FDE?' materialRecruiter screens filter hard on motivation, and generic answers get downgraded. Use the history: "I want the product-discovery version of engineering — embed with one customer, ship the gravel road fast, and feed what I learn back into the product." That's the role described in the industry's own language.
◆ Lock it in
- AAE/FDE = ship products on top of foundation models, not train them.
- The role blends SWE + applied LLM craft + customer/product sense.
- FDE = product discovery: ship the 'gravel road' with one customer; product paves it for the rest.
- FDE leans customer-facing and on-site; AAE leans product engineering — same core skills.
Feynman drill — say it out loudIn 20 seconds, explain to a non-technical friend how a Forward Deployed Engineer is different from a data scientist or an ML researcher.
Step 1 rate your confidence · Step 2 pick your answer
Which best captures the core of the Applied AI / FDE role?
Step 1 — how sure are you?