Agent Deployment Engineer

Sylvan Labs•New York, NY

3d•$160,000 - $230,000•Onsite

About The Position

Sylvan is building autonomous revenue teams to help B2B companies grow and protect revenue from their existing customers by identifying revenue-driving signals and deploying agents that act on them at the exact moment, across every account. We work with customers ranging from fast-growth startups to Fortune 500 enterprises like ServiceNow. Our ambition is to power the top-line of the economy. We are guided by a core principle: no assholes, high mutual respect. We operate at high intensity, but with deep respect. Direct, honest, and no egos — just a shared obsession to win. We’ve raised $11M from world-class investors and strategics including TQ Ventures and HubSpot Ventures. We are a fully in-office company based in New York City. We will only hire A-players and are highly willing to go above-market in ownership, equity, and cash compensation. We want world-class builders who are excited by the opportunity to be part of a category-defining company at the earliest stage. It won’t be easy. We’re going to move fast. But, it’ll be fun. The hardest challenge in post-sales is that every customer's data model is bespoke. Churn means something different at every company. A "healthy customer" is defined differently by every team. Translating those business definitions into verifiable data queries — and proving the signals are right — is where our customers win or lose. That translation is your job. Getting from that to agents taking action on live accounts is the hardest problem in the company, and it is the one that gates everything else. You own that path end to end. The job splits roughly 60% plumbing and translation, 20% signal design, 20% agent engineering. It is customer-facing and it is hands-on: you will run a call with a director of data analytics and a VP of Customer Success back to back, and you will ship the fix yourself afterward.

Requirements

2-5 years of shipping production software or data systems, with at least one role where you owned the customer or stakeholder relationship directly (e.g., analytics or data engineering, forward deployed or solutions engineering, implementation or technical consulting, or similar)
Ships production code independently and agent-driven, at the bar of diagnosing a broken pipeline step at a customer, writing the eval, and shipping the fix without line-by-line review
Warehouse fluency: SQL, dbt, and comfort inside an undocumented five-year-old schema with nobody left to ask
Applied data science literacy across distributions, seasonality, moving averages, validation design, and calibration, with the judgment to know when a result is noise
Credibility with customers under pressure, running a call with a director of data analytics and a VP of Customer Success back to back
Painstaking enough to chase a broken join for three hours because a number looked slightly off
Holds a technical position under pushback from a customer or from us, and changes their mind for evidence rather than for authority

Nice To Haves

Technical co-founder of a company that failed or had a small exit
Technical PM with a real software engineering background
Forward deployed or customer engineering at a data or AI platform company
Analytics engineer or data scientist who moved toward delivery and got tired of research
Have onboarded enterprise customers onto a data product and felt the pain of their warehouse firsthand
PE, finance, or consulting background paired with real software engineering and data science

Responsibilities

Own customer onboarding end to end. Data access scoping, warehouse exploration, and the mapping of a chaotic schema onto the business concepts our ontology needs. You interview their data team and their business owners, and you commit what you learn to the client context repo so the next person inherits it.
Build the signals our agents act on. Signal construction sits on top of the ontology layer and determines whether an agent's judgment is any good. You design the validation that proves a signal is real before it reaches a customer's renewal pipeline.
Diagnose and fix what breaks at a specific customer. When a step in our signal pipeline produces bad output for one account base, you find the cause, write the eval that catches it, and ship the fix. Independently, agent-driven, without someone reviewing every line.
Make the next deployment faster than the last one. Validation frameworks, eval harnesses, mapping accelerators — the internal tooling that turns a bespoke onboarding into a repeatable one.
Own technical trust with the customer and feed the roadmap. Charles owns the commercial relationship; the technical credibility is yours to build and yours to lose. What breaks in your deployments is the primary input to what we build next.