Deployed Engineer at Cognition: Role, Salary, and Interview (2026)
Here is the strangest fact about the Forward Deployed Engineer role at Cognition, the company whose entire product is an AI that writes software: its FDE interview has no coding round. No system design round either. According to a Cognition interviewer’s account aggregated on Exponent, the loop drops both of the rounds that define almost every other engineering interview and replaces them with customer simulations — an executive pitch, a timed case-study call, a project presentation to a skeptical panel. The company that built Devin, “the first AI software engineer,” has decided that the last mile of its business is a human problem, not a coding one. That decision tells you almost everything about what this seat is.
The company, briefly
Cognition is not a lab renting out a foundation model. It’s an applied-AI company built around a single product — Devin, the autonomous coding agent — and it has grown at a pace that is hard to overstate. Devin scaled from roughly $1M in ARR in September 2024 to about $73M by June 2025. Then in July 2025 Cognition acquired Windsurf — the AI IDE — after Google hired away Windsurf’s CEO and OpenAI’s acquisition bid collapsed, absorbing roughly $82M in additional ARR, 350-plus enterprise customers, and the engineering team. By May 2026 the combined company was estimated at around $492M in ARR (per Sacra) and had raised over $1 billion at roughly a $26 billion valuation, up from $10.2B just eight months earlier. Founded by Scott Wu and a talent-dense team drawn from Scale AI, Palantir, Cursor, Waymo, and Google DeepMind, Cognition is the kind of company where the org chart is still being drawn — which is exactly why the customer-facing seat matters.
What the “Deployed Engineer” actually does
Cognition doesn’t call the role a Forward Deployed Engineer. On its careers board it’s the Deployed Engineer (DE) — and as Exponent notes, the work matches what the rest of the industry calls an FDE. The DE owns “all technical aspects of the customer lifecycle, from deployment through activation,” according to the live posting: leading demos and pilots, connecting Devin and Windsurf into real production environments, wiring the product into services like Jira, Linear, and internal databases, and handling the on-premise edge cases that documentation never covers. They join early customer calls — including first conversations with CTOs — as the technical voice in the room, and they’re expected to ship: the posting explicitly asks for “a bias towards implementing requested features yourself.” Travel runs 25–50%.
There’s a second variant worth flagging: a Deployed Engineer, Federal seat based in Washington, DC and San Francisco. That’s a meaningful signal. When an AI company spins up a dedicated federal deployment track, it’s telling you the government pipeline is real enough to staff against — and federal work carries its own clearance, on-premise, and compliance texture that the commercial DE role doesn’t.
Compensation
Cognition posts base bands by track. Reading them off the company’s own Ashby postings (as compiled by Exponent), the picture is:
| Role | Base band (2026) |
|---|---|
| Deployed Engineer (San Francisco) | $98K–$170K |
| Deployed Engineer, Federal (San Francisco) | $130K–$225K |
| Equity | Private options at ~$26B valuation |
Two honest notes. First, these are base salary only — the real story at a company that has more than doubled its valuation inside a year is the equity, and options struck against a $26B mark are no longer cheap, so the recruiter conversation about preferred-share price and refresh grants matters more here than the base band. Second, the DE base sits below the frontier-lab FDE numbers in our salary guide — the OpenAI and Sierra versions of this job pay more in cash. What Cognition is selling instead is equity torque and proximity to one of the fastest-growing products in software. That’s a different trade, and it’s worth being clear-eyed about which one you’re taking.
The interview: an FDE loop with the code removed
This is where Cognition is genuinely distinctive. Based on the interviewer account on Exponent, the loop runs roughly like this:
| Stage | What it tests |
|---|---|
| Recruiter screen | Communication, motivation for a customer-facing pivot |
| Technical take-home | A hands-on exercise inside Devin — minimal coding, mostly proving you can drive the product like a customer |
| Project presentation | ~45 min walkthrough of past work, explained so a non-technical panel grasps the impact |
| Leadership 1:1s | Behavioral conversations with VPs and engineering leaders |
| Executive pitch | A simulated pitch of the product to a panel role-playing company executives |
| Case-study customer call | A timed prep window, then a live simulated call with a fictional customer |
Read the pattern. Every stage is about whether you can explain why technical work matters to someone who doesn’t share your vocabulary, and whether you keep your composure when a panelist plays a difficult customer and asks something you can’t fully answer. The take-home isn’t a LeetCode gauntlet — it’s setting up and running a real process inside Devin, which quietly assumes you’re already fluent in the developer workflows the product automates. The company’s stated failure mode is telling: the most common reason strong engineers get rejected, per the guide, is “technical fluency without customer communication” — walking through how you built something without ever saying why it mattered.
There’s a logic to removing the coding round that only makes sense at this specific company. Cognition sells the machine that does the coding. What it can’t automate — yet — is the engineer sitting across from a skeptical CTO, translating an agent’s output into business value and earning the trust that gets Devin adopted rather than shelved. So that’s the only thing it interviews for.
Who should apply
The DE seat rewards a specific person: an engineer with a real technical foundation who has quietly decided the interesting problem is adoption, not architecture. If you like the Solutions Engineer instinct — customer in the room, product in production, trust as the deliverable — but you want to keep your hands on the system, this is a clean expression of that. Two caveats. Cognition has publicly described an “extreme performance culture” including long hours and weekend work, and the postings ask for “high tolerance for ambiguity, intensity, and sustained effort” — take that at face value. And if you’re a pure builder who wants a coding-heavy loop and a coding-heavy job, note that Cognition runs a separate, more code-intensive software-engineer process; the DE role is not that. Apply to the one that matches the work you actually want.
Cognition’s Deployed Engineer roles — commercial and Federal — are live on our job board now. Get the weekly digest to track new FDE openings as they post.