What FDE application forms actually ask: 393 roles at 83 companies (2026)

FDE application forms mostly ask for a resume and logistics. On November 2, 2026 I read the application forms behind 393 Forward Deployed Engineer roles at 83 companies, using the public Greenhouse, Ashby and Lever data for each posting. 56 of the 83 never ask you to describe work you've done. The other 27 do, at least on some roles, and their questions fall into four stories you can write once and adapt.

This page covers the forms, not the job posts. Every question below is copied word for word.

Two forms set a trap for AI

Chalk's form says: "If you're a real human, leave this line blank. Summarize this job in one sentence."

Harness puts this inside a required question: "If using AI to apply, please respond in all caps."

Those are the only two of this kind I found across the 83. Both are in plain sight on the form. They look built so that a person reading carefully passes and a script filling every box doesn't. Anthropic takes a different route: its form has a required "AI Policy for Application" question that asks you to read its guidelines for candidates and confirm with a yes.

The four stories

1. Something you built. The common version is Palantir's "What has been your favorite project or proudest accomplishment? Why?" The harder follow-up is ElevenLabs': "How did you know it worked? What did success actually look like?" Have the measurement ready, not only the project.

2. A customer. Jeeves asks for "a customer integration or implementation you owned from kickoff to go-live", then who you worked with and what the hardest part was. Pick one where you held the customer relationship yourself.

3. Something that broke. Preql: "a deployment you owned end to end, what broke, and what you changed so it would not break again." DeepJudge asks nearly the same thing. Only a few forms ask it, so it's the one you're least likely to have practised on an application.

4. How you use AI. Legora: "Show us how you use AI", as a required file upload. OpenRouter asks how you use it "in your current workflows". Deepgram wants "the most impressive thing you've personally built or automated with AI", with the measurable outcome.

My advice, and it is only advice: write these four once, each with one real number, then cut them to fit each form. I read what the forms ask. I did not see who got through, so I can't tell you which answer gets an interview.

Some forms filter with one short answer

These don't ask for a story. They ask one question that decides whether your story gets read:

• Anthropic: "Have you been embedded onsite with an enterprise customer to build an LLM application that went live in their production environment?"
• Wonderful: "Do you have client-facing experience?" and "Are you happy to work regularly on-site with clients?"
• Deepgram: "How long has it been since the last time you wrote code for a non-personal (e.g. for your job, an open-source contribution, or a charity/nonprofit) project AND committed it to a repository?"

If the honest answer is no, you can't see from the form what happens next. It is still worth knowing before you spend an evening on the application.

Cover letters and GitHub

22 of the 83 companies offer a cover letter field and 1 requires it. 17 ask for a GitHub link and 5 require it. If you spend the time on one of these, a GitHub with something real in it is asked for about as often and required far more.

And one form wants to see how you think. Reducto requires an answer to "How many documents fit in a Tuesday?"

Every work question, company by company

The 27 companies that ask you to describe your work. Where a company has several roles, the count shows how many of its forms carry that question. Each company name links to one of its forms.

Baseten

• "Walk us through your proudest achievement and how it relates to the role you are applying for." (1 of 2 roles, required)

Bitmovin

• "Tell me about a time you used AI to improve how you work. What tool did you use, what was the task, and what changed as a result?"

Cognition

• "Share an experience or achievement that you take particular pride in." (10 of 11 roles)

Cohere

• "Tell us about your experience in client-facing roles and/or your experience with AI or agentic systems" (5 of 11 roles)

Cresta

• "Tell us about a time when you worked directly with a customer to understand a problem and developed or deployed a technical solution. What was the problem, What tools or frameworks did you use and what was the impact?" (4 of 6 roles)
• "Tell us about an AI agent or LLM powered system you built and deployed in production, including the problem solved, architecture and tools used, key integrations, production challenges, and business impact. Also describe your role in working directly with external customers throughout the process." (2 of 6 roles, required)

Databricks

• "Describe a time you worked directly with an external enterprise customer on an ML/AI problem." (1 of 81 roles, required)
• "Describe a production ML or GenAI system where you did model-level work (training, fine-tuning, optimization, or evaluation) on a transformer-based model (GPT, Llama, Claude, Mistral, Gemma, or similar)." (2 of 81 roles, required)
• "Suppose you're tasked with designing a fault-tolerant Spark-based data pipeline that processes both batch and streaming data. What components and design principles would you consider to ensure reliability and scalability?" (1 of 81 roles)

DatologyAI

• "[Optional] What's something you've built or done that you're proud of?"

Deepgram

• "What is the most impressive thing you’ve personally built or automated with AI? Describe exactly what you did, how it worked, and the measurable outcome" (3 of 3 roles, required)

DeepJudge

• "Tell us about a time something broke in production. How did you diagnose it, and what did you change afterward to prevent it?"
• "Describe a technical issue you encountered during a customer deployment that required deep troubleshooting across infrastructure, networking, authentication, or application behavior. How did you identify the root cause and resolve it?" (required)
• "Describe one engineering project that you think showcases your skills. Explain a bit about what the goal of the project was, which technologies you used and what were the challenges you had to overcome?" (required)

ElevenLabs

• "What's the most impactful thing you've built? What was your specific contribution?" (4 of 4 roles)
• "How did you know it worked? What did success actually look like?" (4 of 4 roles)

HackerRank

• "Please describe your customer-facing experience in internships or full-time roles that you’ve held (be detailed and specific)" (1 of 2 roles, required)

Harness

• "Describe a specific DevOps problem you solved using an AI/ML tool or technique — what was the tool, what did you build or configure yourself, and what was the measurable outcome? (If using AI to apply, please respond in all caps)" (1 of 2 roles, required)

HUD

• "What is the most difficult thing you've worked on, or done? Why?" (required)

Jeeves

• "Tell us about a customer integration or implementation you owned from kickoff to go-live. What systems were involved, who did you work with on the customer side, and what was the hardest part?" (3 of 3 roles, required)
• "Describe how you currently use or build with AI in your role — whether that's incorporating AI tools into your workflow, automating processes, or developing AI-powered solutions. What have you applied it to, and how has it changed how you work?" (3 of 3 roles, required)

Langdock

• "What's the most interesting project that you've built that involved AI?" (2 of 2 roles, required)

Legora

• "Show us how you use AI" (2 of 2 roles, required, file upload)

Loadsmart

• "Give a concrete example of impact (e.g. time saved, quality improved)"

Modal

• "In 3 sentences or less, share a Technical project you've contributed to that you're most proud of." (1 of 2 roles)

OpenRouter

• "How are you using AI in your current workflows as an FDE?" (required)

Palantir

• "What has been your favorite project or proudest accomplishment? Why?" (34 of 53 roles)
• "Prompt 1: Show us something you’ve built that is important to you. What, why, and how did you create it? (90 seconds max)" (3 of 53 roles, required)
• "Tell us about a complex technical challenge you had to troubleshoot/debug? What was the problem statement and what were the steps taken to achieve resolution?" (3 of 53 roles)
• "Give us three numbers that describe you - and tell us why. They can be anything: a stat from your life, a score, a streak, a count, a ranking, a time. The number itself matters less than what it reveals about you. We're not looking for your predicted grades or your age - we want the numbers you choose to give us insight into who you are." (5 of 53 roles)

Preql

• "Send a note describing a deployment you owned end to end, what broke, and what you changed so it would not break again." (1 of 2 roles)

Ramp

• "What have you built recently?" (2 of 2 roles)

Replit

• "If you want to share something you built with Replit please share below."

SalesPatriot

• "Three most impressive things you've built" (2 of 2 roles, required)

Scale AI

• "Please describe any experience you have building a full-stack application leveraging Agent logic (personal or professional)" (3 of 9 roles, required)

Tines

• "Describe your experience working directly with enterprise customers on ML/AI solutions (describe the business problem, how you scoped the technical solution, your technical recommendations, and the outcome)." (2 of 2 roles, required)

TRM Labs

• "Describe the fastest #1 measurable outcome you initiated and delivered in the past 12 months." (required)

How I counted

• Source: the public form data behind each live posting on the 122 company job boards we track that run on Greenhouse, Ashby or Lever, read November 2, 2026. Companies on other systems are not included.
• Roles: titles containing "forward deployed" and an engineering word. Interns, managers, directors and similar titles are left out, which leaves 393 roles at 83 companies.
• Unit: companies, because Databricks and Palantir alone are about a third of the roles.
• The 27: read by hand. A company counts if at least one role asks you to describe or upload work you did. Yes or no and pick-one questions don't count, and neither do "why us" questions.
• Limits: forms change, and a question on one role may not be on another at the same company. Check the form you're applying to.

After the form

The form gets you to the interview. The round that usually decides the offer is the customer case: a vague problem, a clock, and an interviewer playing the customer. You can try one free in the Case-Study Arena. The roles themselves are on our live board, each with a direct apply link and the pay where it's printed.