Forward Deployed Engineer at Decagon: Role, Salary, and Interview (2026)
Decagon sells customer-support agents on a pricing model that makes its own Forward Deployed Engineers the most important people in the building. Per its December 2024 pricing post, it charges per conversation, or a higher rate per resolution (nothing if the agent escalates). Revenue only exists if the agent actually works inside a customer’s messy ticketing stack. That is a deployment problem, and Decagon hires FDEs to solve it.
The company is growing fast enough to matter. Decagon’s annualized revenue was about $10 million at the end of 2024 (The Information), and in August 2026 the company told Newcomer it had passed $100 million. That is a company disclosure, not an audited figure. Headcount was 300+ at its March 2026 employee tender.
The company behind the seat
Founded in August 2023 by Jesse Zhang (CEO) and Ashwin Sreenivas (CTO), Decagon has raised roughly $481 million across five priced rounds. The most recent was a $250 million Series D in January 2026 at a $4.5 billion valuation, led by Coatue and Index Ventures, roughly triple the $1.5 billion mark from June 2025. Reported customers include Avis Budget, Hertz, Block, Affirm, Chime, Varo, Deutsche Telekom and Grubhub. The company claims average deflection rates above 80%.
What the job actually is
The posting we could read is Forward Deployed Engineer, Agent Builder, based in San Francisco or New York, in a team Decagon calls the Agent Builder Org, a “technical delivery team” under Post Sales Operations. Note where it sits: not in Engineering, and not in Sales. It is a delivery function.
The work, per the posting:
- Run enterprise agent builds from scoping through launch and iteration
- Write and maintain the artifacts that define agent behavior (Decagon calls them AOPs, or Agent Operating Procedures), then tune for quality and reliability
- Configure and validate guardrails
- Set up and test integrations (ticketing systems, internal tools) and build whatever tooling those need
- Agree success criteria with senior customer technical stakeholders
- Feed customer insights back to Engineering
A sister posting in Australia, titled Agent Deployment Engineer, adds detail worth reading: debugging “probabilistic and technical failures” through root-cause analysis, and building evaluation and regression-testing frameworks. Neither posting mentions travel.
Two things stand out. First, this is much closer to a configuration-and-evaluation job than to the Palantir model of writing software on site. The core artifact is an agent specification and the test harness around it. Second, the bar for entry is seniority, not AI experience: the requirement is 5+ years in a customer-facing technical role (solutions, forward-deployed or implementation engineering), and LLM or agent experience is only “nice to have.” If you are a strong implementation engineer from a Salesforce or Zendesk ecosystem, this is a door. Compare with Sierra, which runs a similar agent-deployment model.
The title keeps changing
If you search for this role you will find it under several names. Decagon’s own board shows “Forward Deployed Engineer, Agent Builder” and “Agent Deployment Engineer,” and third-party job aggregators list apparently equivalent seats as “Customer Engineer, Agent Builder,” including a French-speaking variant. We could confirm the first two on primary postings and could not open the third at source, so treat that one as unverified. The pattern still matters for your search: if you filter only for “Forward Deployed Engineer,” you will miss part of the pipeline.
Compensation
| Posting | Location | Posted pay |
|---|---|---|
| FDE, Agent Builder | SF / NYC | $175K–$230K base + equity |
| Agent Deployment Engineer | Australia | A$200K–A$270K (header shows A$213K–A$266K) incl. equity |
That $175K–$230K is base, and it is the number to anchor on. It sits well below the $350K–$550K total-comp bands the frontier labs post, and below Ramp’s posted bands, which we covered in our Ramp profile. The tradeoff is equity in a company valued at $4.5 billion with $100 million in reported run-rate. Ask for the strike price, the latest 409A and the refresh policy before comparing it with a lab offer. The Australia posting is inconsistent between its header and its body, a reminder to confirm the band with the recruiter.
The interview
Decagon has not published an FDE-specific loop, and the third-party guides we found explicitly don’t cover it. What follows is the engineering loop as reported by candidates and a third-party guide (techinterview.org, updated October 2026); expect your track to overlap but verify with the recruiter.
| Stage | Format | Reported content |
|---|---|---|
| Recruiter / hiring call | Short | Background, motivation |
| Coding | 60 min | Practical build tasks; one candidate reported a shopping cart with undo and redo |
| System design | 60 min | LLM-oriented; one candidate was asked to design security for a chatbot product |
| Project deep dive | 60 min | Several layers deep on something you drove, including rejected alternatives |
| Behavioral / culture | Varies | A project that went wrong; how you use AI |
The guide puts the whole cycle at 3–4 weeks. The most useful prep is the design question: build an agent that resolves a ticket end to end, covering intent, tool calls, escalation, guardrails and ambiguous requests. Pair that with an answer to “how do you define resolved,” because Decagon’s own pricing turns on it. One candidate reported rejection straight after the culture round, so don’t treat that stage as a formality.
Who should apply
Apply if you have five-plus years of implementation or solutions engineering and want to be close to a product with measurable outcomes. Skip it if you want travel-heavy, on-site software building, or a frontier-lab compensation package. Decagon describes itself as an in-office company.
Browse current forward-deployed openings on the job board, or get the weekly digest.
By A. George. Postings change; the SF/NYC posting was showing as no longer accepting applications on one mirror at time of writing.