Marketing Operations Engineer

Job Description

Job Description

About The Role

AI is the new electricity. Founded and led by Andrew Ng, DeepLearning.AI is on a mission to help everyone build with AI. Ten million people learn with us, and nearly all of them reach us through one system: HubSpot at the center, Customer.io sending on top of it, a warehouse and BI layer behind it, analytics feeding both, and a marketing site in front of it all.

You'll own that system — the architecture, the standards, and the numbers that come out of it. You design how the stack fits together, set the conventions everyone builds on, and direct the contractors who execute against them. You're not the person running a monthly hygiene cycle by hand; you're the person who defines what good looks like, builds the system that does it, and catches it when it doesn't.

This role exists because we believe the next unlock for our team isn't more headcount doing manual work – it's state-of-the-art agentic techniques doing the work with just a bit of human involvement. As a natural extension of the work, you'll often be the most credible source on our team for what AI tooling can actually do today, which gives us a built-in story for our own content.

When someone asks "is this number right" or "why isn't this attributing," you're who they ask. When someone asks how we define a new versus returning user, your answer is the answer, and it's written down.

You won't be writing copy, building sends, or running the publishing calendar. You'll be building the platform for the people who do.

You will report to the Head of Marketing and work in an office collaborating with some of the most forward-thinking AI engineers building the next generation of technology.

What you'll do

Build the automation layer

• Identify which marketing workflows are highest-leverage to automate first, in partnership with the Lifecycle Marketing Manager and Marketing Operations Coordinator.

• Design and build agentic workflows against the systems you own: campaign QA, data validation, reconciliation, segmentation logic, and the reporting the team does by hand today. The weekly performance report should pull itself.

• Build agents that check the system continuously rather than waiting for someone to notice — a sync that stopped, a property arriving null, two dashboards that stopped agreeing.

• Design for verification, not just output. An agent reporting confidently on bad data is worse than no agent, so the evals ship with the workflow.

• Build tooling that makes routine work self-service, so tactical requests go down over time rather than up.

• Continuously evaluate new agentic and AI tooling (Claude, other LLM APIs, automation platforms) and prototype how it applies to our specific funnel problems.

Own the stack

• Own HubSpot as our system of record: object model, property architecture, workflows, permissions, reporting infrastructure, contact tiers and what they cost us.

• Own the integration layer connecting HubSpot, Customer.io, the warehouse, PostHog, Stripe, and our mobile app, including the HubSpot–Customer.io seam. Which system holds truth for an attribute, how it gets there, what happens when they disagree. Write the middleware where native connectors fall short, and the monitoring that catches a sync failing silently.

• Own the contact database and the schemas underneath it: property governance, lifecycle stages, dedupe logic, and the segmentation architecture campaign owners build on. Design events and properties that survive our next three launches.

• Own deliverability infrastructure including authentication, domain reputation, bounce and complaint rates – instrumented so problems surface on their own.

• Be the liaison to engineering for the marketing site: know what marketing can change and what needs a developer, own the technical side of new page builds, and file the kind of ticket that gets picked up on the first read.

Own tracking and measurement

• Own UTM conventions, tag deployment, and server-side conversion tracking, in partnership with engineering. Documented so that people follow them.

• Instrument new surfaces before they launch. When a product, page, or app build ships, tracking is live on day one because you scoped it during planning.

• Be the source of truth for marketing performance. Own the definitions, not just the dashboards including enrollments, memberships, revenue (MRR, ARR, churn, LTV), app installs and attribution, campaign performance, and write the queries behind them with Data Engineering where it goes deeper.

• Flag the caveats before someone builds a decision on them. A channel group with three months of history is not a trend, and you're the one who says so.

Own experimentation infrastructure

• Build the infrastructure that lets the team test rigorously without engineering as a bottleneck: assignment, holdouts, sample sizing, and clean test structure so one variable moves at a time.

• Support experimentation across lifecycle, acquisition, and paid. Give paid social a measurement foundation — audience and geography definitions, conversion tracking, and reporting that says what actually worked.

Direct the work you don't do yourself

• Scope, brief, and review contractor work. Write specs clear enough that execution doesn't come back wrong, and know the work well enough to catch it when it does.

• Define recurring processes, automate what can be automated, and hand off the rest with the standard attached.

Dogfood & Content (Secondary)

• Where it's a natural byproduct of the automation work, document exciting or interesting builds for use in our content across our newsletters, YouTube accounts, events, or technical blogs. Note this supports the content engine but is not a primary deliverable or goal.

• Partner with Developer Relations when a build is interesting enough to become a public case study or tutorial.

What you should have

• AI-native, default to using AI-assisted coding and building automations in everything you do. Appetite, passion for and proven record of learning and experimenting with the newest AI engineering best practices.

• 5+ years building or owning marketing systems — marketing engineering, marketing ops, growth engineering, or a software role embedded with a marketing team. Somewhere you owned the stack rather than just used it.

• Real engineering ability. Production JavaScript or Python, comfortable against REST APIs and webhooks, at home in Git. You've shipped things other people depend on.

• An automation instinct. You see a recurring manual task as a system to be built, and you've built them. You use LLMs daily and have a real point of view on what they can and can't be trusted with.

• Deep experience administering a marketing automation or CRM platform — HubSpot, Customer.io, Marketo, SFMC, Braze, or comparable. If it isn't HubSpot, tell us how you'd get dangerous in it inside 60 days.

• Comfort with data. You write and debug SQL against a warehouse, build in a BI tool like Metabase, and use GA4 without treating its numbers as gospel.

• Real understanding of tracking and attribution: UTM conventions, tag management, campaign structure, server-side conversion tracking, and where attribution commonly goes wrong.

• An instinct for verification. You reconcile against a second source before publishing, and you notice when a dashboard has been quietly wrong for a month.

• A track record of running something end to end — a migration, a platform build, a consolidation — including the deprecation and the stakeholder work.

• A documentation habit. You leave definitions and conventions written down so the next person doesn't reverse-engineer them.

• On-site reliability. Mountain View, five days a week.

Bonus if you have

• Hands-on experience building with LLM APIs, agentic patterns, or eval frameworks

• Experience scoping and reviewing contractor or vendor technical work

• Product analytics and event schema design (PostHog, Amplitude, Mixpanel)

• Mobile attribution (AppsFlyer or comparable)

• Stripe, Jira, Semrush, or tag management experience

• Headless or publishing CMS platforms and deployment tooling like Vercel

• Fast-paced startup or technical environment

• Experience writing technical content or documentation aimed at a developer audience

• Familiarity with the AI/ML or technical education landscape

What Success Looks Like

In 30 days you've mapped the stack, found where the numbers disagree, and built your first agent to help make things work, with a documented build that can double as content. In 90 days the definitions are documented, the weekly report pulls itself, and contractors are working against specs you wrote. In 6 months you own a function the team trusts by default, with enough automation that your time goes to the next problem instead of maintaining the last one.

We may use artificial intelligence (AI) tools to support parts of the hiring process, such as reviewing applications, analyzing resumes, or assessing responses and identifying potential inconsistencies or verification signals in application materials based on available information. These tools assist our recruitment team but do not replace human judgment. Final hiring decisions are ultimately made by humans. If you would like more information about how your data is processed, please contact us.