Decagon Engineering Jobs: How to Apply in 2026 (And Which of the 55 "Engineer" Roles You Actually Want)

We built Standout because the application-driven job search hides the information that decides the outcome, and Decagon's job board is a clean example. Every page ranking for this query tells you Decagon is hiring engineers and points you at the same button. None of them tells you that the word "engineer" on that board is doing two completely different jobs, or that Decagon's own posted compensation line gives away which one you are looking at.

Decagon is a conversational AI company that raised $250M at a $4.5B valuation in January 2026. Its job board lists 137 open roles, 55 of them with "Engineer" in the title. Those 55 are two different careers, not one: 29 core product and research reqs, 25 customer-facing deployment reqs. The posted comp line tells you which is which.

Every req, band, team and location number in this article was pulled directly from Decagon's own public job board API on August 24, 2026, not from an aggregator.

The board says "Engineer" 55 times and means two different jobs

Fifty-five of Decagon's 137 open postings carry an engineering title. Twenty-nine of them sit on core product, platform and research teams: Agent SWE (8), Research (6), Agent Orchestration (4), Infrastructure (4), Security (4) and Enterprise Product (3). Twenty-five sit on the customer-facing side: Solutions Engineering (15) and Agent Deployment Engineering (10). Three more are neither, and they are the ones that catch people out. Two are IT Engineer reqs at $96K to $140K, and one is a GTM Analytics Engineer on Revenue Operations at $190K to $230K (Source: Decagon job board API).

Counts include non-US reqs; the bands shown are the US ones. London, Toronto, Australia and EU reqs post in local currency, and five Senior Solutions Engineer reqs post no band at all (Source: Decagon job board API).

The hot take: you do not need to read a single job description to tell the two tracks apart, because Decagon's own comp line gives it away. Every Solutions Engineering req says "Offers Commission." No core engineering req does (Source: Decagon job board API). Commission on an engineering title means the role is measured on customer outcomes, and that is a different job from the one you think you are applying to. Read the comp line before you read the JD.

None of this makes the deployment track the lesser career. Solutions Architect (Salesforce) posts $300K to $375K, and that $300K floor is the highest base floor of any engineering req on Decagon's US board, higher than the $280K floor on Engineering Manager and $100K above the floor every core IC req starts from (Source: Decagon job board API). It is a different job, and for a specific kind of engineer it is the better one.

Twenty of Decagon's 24 US core IC engineering reqs post the identical band: $200K to $400K

Decagon has 27 core engineering reqs across San Francisco and New York. Three are Engineering Manager reqs. Of the 24 individual-contributor reqs left, 20 post the identical $200K to $400K base plus equity, and the other four, all on Security, post $200K to $330K. Two bands, 24 reqs. Senior Software Engineer, Staff Software Engineer, Research Engineer, Senior Research Engineer, Staff Research Engineer, Agent Platform, Cloud Infrastructure, Data Infrastructure, Enterprise Product and Voice Agent all sit inside the same $200K to $400K number (Source: Decagon job board API).

All reqs above state "Offers Equity." London core engineering posts separately, at GBP 200K to 300K for Senior Research Engineer and GBP 125K to 180K for Software Engineer, Agents (Source: Decagon job board API).

Two things fall out of that table. Security is the only core team that prices differently, at $200K to $330K across all four of its reqs. And Engineering Manager posts $280K to $430K, a floor $80K above every core IC req, which is an unusual inversion at a company this stage. Only one engineering req on the whole US board starts higher, and it is a customer-facing one (Source: Decagon job board API).

The hot take: a single $200K-wide band shared by "Research Engineer" and "Staff Software Engineer" is not a pay scale, it is a placeholder. Your level is priced after the company meets you, not by the req you clicked. That cuts both ways and the honest version says so. It is real upside if you arrive with someone credible already vouching for where you land. It is real downside if you arrive as application number N, because the default anchor in that conversation is the bottom of the band, and nothing on the req argues you out of it. Decide before you apply which of those two candidates you are going to be.

Why the deployment org is nearly as big as the product org

Decagon's product is an agent that closes customer support tickets for enterprises without a human behind it. In Decagon's own Duolingo case study, deflection went from roughly 30% with the previous vendor to 80%, live one month after the first conversations (Source: Decagon). That number does not get shipped by writing the model. It gets shipped by an engineer sitting inside the customer's stack, wiring an agent to real systems and finding out where it breaks against real traffic. Decagon says it added more than 100 new enterprise logos in 2025 (Source: Decagon), which is the demand side of exactly that work.

So the hiring plan has the shape of the product. An enterprise deployment company hires deployment engineers at close to parity with product engineers, and a candidate who reads 25 field engineering reqs as evidence of overhiring has it backwards. That is the company telling you where the hard problems are.

From the matches we run at Standout, this second engineering track is now growing faster than the first at the companies we represent candidates into, and candidates keep filtering it out of their search because the title has the wrong word in it. If you have ever been the engineer who got dropped into a customer's environment and enjoyed it, stop screening these reqs out.

This is a San Francisco job, and it is not a remote one

San Francisco 73, New York City 24, London 17, then single digits across Australia, Toronto, Amsterdam, Austin, Dallas, Sweden, Austria, France, Germany and Denmark. 104 of the 137 reqs are US-based. Four postings say Remote, and all four are go-to-market: two Strategic Account Directors and two Strategic Solutions Engineers. Zero core engineering reqs are remote (Source: Decagon job board API).

If you are remote-only, this section is where the article ends for you. There is nothing on this board to apply to.

Headcount reportedly went from roughly 100 to 300+ in about 18 months across San Francisco, New York and London, per a secondary source dated April 16, 2026 (Source: JobsByCulture). Decagon publishes no engineering-specific headcount, and neither does anyone else, so we are not giving you a ratio of engineers to reqs.

One honest limit on our own usefulness here: 33 of these 137 reqs sit outside the US, 17 of them in London (Source: Decagon job board API). Standout places in the US only. On those 33 we are no help at all, and you should go straight to the board.

The loop, and the round that nobody, including the people who sat it, can explain

The most consequential round in Decagon's loop is the one no prep site can prepare you for, because Decagon has not said what it grades.

An anonymous candidate on Blind described passing the phone screen and being invited to an onsite that included an "AI Coding" round built around shipping an app using AI, and posted at the time that they had never sat an AI coding interview and did not know what to expect. Asked afterward how it went, the same poster wrote: "It was pretty straightforward. Nothing to prepare for, be ready to use AI in your personal setup," then added that they did not understand what the interviewers were looking for and that the round felt low-signal (Source: Blind).

Most companies are still arguing about whether to ban AI assistants in interviews. Decagon has walked past that argument and made bring-your-own-AI an actual round, without publishing a rubric for it. The round is not unfair. It is unspecified, which is a different problem and a worse one to walk into cold. The only preparation that survives contact with it is arriving with a setup you already drive fluently every day, not one you installed the night before to look AI-native.

The rest of the loop, from a third-party guide rather than anything Decagon publishes: a 30-minute recruiter screen, a 60-minute coding pair in TypeScript or Python often with a concurrency or LLM-tooling flavor, a 60-minute system design round on building an agent that resolves tickets and where it breaks, a 60-minute past-project deep dive, and a 45-minute behavioral round, at three to four weeks start to finish (Source: techinterview.org). Glassdoor respondents in Software Engineer roles rate the interview 3.1 out of 5 on difficulty with 76% reporting a positive experience (Source: Glassdoor). The most immediately useful detail comes from Blind again: a respondent reported the recruiter saying Python is preferred, JavaScript seemed acceptable, and C++ did not (Source: Blind).

Three things we will not tell you, because nobody can source them. The accounts disagree on the coding screen itself, with the guide describing something practical and non-algorithmic and a first-hand Blind respondent describing Leetcode arrays and stacks (Sources: techinterview.org, Blind), and we are not picking a winner. Nobody publishes what the AI round grades, and we are not going to guess a rubric on your behalf. And the three-to-four-week figure covers the loop once it has started; no public source says how long an application sits in the queue before it starts, so any application-to-first-call number you find for this company is unsourced, ours included.

On visas: Scoutify reports H-1B sponsorship at Decagon as confirmed, with 5+ USCIS approvals documented (Source: Scoutify).

Everything published about this company is reconstructed from the outside

Give Decagon this first: there are no special-purpose vehicles named for it on SEC EDGAR, which is cleaner than several comparable AI companies where pre-IPO share vehicles show up in the record before the operating company does. Decagon AI, Inc. files as CIK 0002043374 out of Delaware, with exactly one filing in its entire history, a Form D dated December 23, 2024, and no later filing for the January 2026 round (Source: SEC EDGAR). A full-text EDGAR search for "Decagon" returns 67 hits, and almost all of them belong to Decagon Asset Management LLP, an unrelated fund filing Schedule 13Gs across dozens of SPACs (Source: SEC EDGAR).

Now the part that affects you. Fast AI Jobs showed 111 open roles. Scoutify showed 141. Decagon's own board returned 137 when queried directly on August 24, 2026 (Source: Decagon job board API). Meanwhile Decagon's careers page publishes mission, values, benefits, team profiles and global hiring, and publishes no interview process, no salary ranges, no required experience levels and no guidance on how to differentiate an application (Source: Decagon careers). The layer between you and this company is rebuilding the board from the outside and missing by up to 30 reqs, while the company itself withholds the three facts that would change how you apply.

Who should actually apply, and how

The mechanics, in order. Go to decagon.ai/careers and into the Ashby board directly rather than through any aggregator, because the aggregators are wrong by up to 30 reqs. Read the compensation line before the job description and use "Offers Commission" as your track tell (Source: Decagon job board API). Pick one track and one req on it. And if you are not in or moving to San Francisco or New York, settle the relocation question before you start a five-stage loop, not after.

One last thing, and it comes straight out of the comp table above. When 20 of 24 reqs share a single $200K-wide band, your number is not set by the posting. It is set in a conversation, and the anchor in that conversation depends entirely on how you arrived. That is the whole argument for having an agent instead of an application. Decagon is exactly the company profile we represent people into: US tech, this stage, hiring hard. Standout matches you to companies like it, and when you say yes, we introduce you to the founder directly instead of adding you to a queue. It is free for candidates. Read how Standout's matching works before you open another req. If you have already decided the board is good enough, read how Standout compares to job boards first.

And to repeat the concession, because it matters more than the pitch: on the 33 non-US reqs on this board, we cannot help you. Go apply.

FAQ

How do you apply for engineering jobs at Decagon?

Go to decagon.ai/careers, which routes to Decagon's Ashby board, and apply on the req itself rather than through an aggregator. Aggregator counts for this company ranged from 111 to 141 against a live board of 137 on August 24, 2026 (Source: Decagon job board API). Decagon publishes no application guidance of its own, so the req page is the only authoritative source you have (Source: Decagon careers).

How much do Decagon engineers get paid?

Twenty of Decagon's 24 US core individual-contributor engineering reqs post an identical $200K to $400K base plus equity, from Research Engineer through Staff Software Engineer. The other four, all on Security, post $200K to $330K, and Engineering Manager posts $280K to $430K. Customer-facing engineering posts differently: Agent Deployment Engineer at $175K to $230K, Senior Solutions Engineer at $200K to $280K with commission, and Solutions Architect (Salesforce) at $300K to $375K with commission (Source: Decagon job board API).

What is the difference between a Software Engineer and an Agent Deployment Engineer at Decagon?

A Software Engineer builds Decagon's product and platform, on one of 29 core reqs across Agent SWE, Research, Agent Orchestration, Infrastructure, Security and Enterprise Product. An Agent Deployment Engineer is one of 25 customer-facing engineering roles that take the agent into an enterprise customer's stack and make it work against real traffic. The fastest tell is the posted comp line: commission appears on the customer-facing roles and never on the core ones (Source: Decagon job board API).

Does Decagon hire remote engineers?

No. Zero of Decagon's 137 open postings are remote engineering roles, and all four Remote postings on the board are go-to-market. Engineering is concentrated in San Francisco, New York and London (Source: Decagon job board API).

How hard is Decagon's engineering interview?

Glassdoor respondents in Software Engineer roles rate it 3.1 out of 5 on difficulty with 76% reporting a positive experience (Source: Glassdoor), across a reported five-stage loop of recruiter screen, coding pair, system design, project deep dive and behavioral, at three to four weeks (Source: techinterview.org). The wildcard is an "AI Coding" round that one anonymous candidate described as requiring your own AI setup and offering nothing to prepare for (Source: Blind). No public source says how long an application waits before the loop starts, so treat any application-to-first-call figure you see as unsourced.

---

You do not have to be application number N. Standout is a talent agent for US tech professionals. Build one profile, get matched to companies at Decagon's stage, and get introduced to the founder directly instead of joining a queue. First matches arrive within a few hours. Free for candidates.