Forward Deployed Engineer, Applied AI

About Snowplow Snowplow is the real-time customer context layer that collects, validates, enriches, and delivers behavioral data for advanced analytics, ML, and AI agent decisioning. Snowplow's event tracking is leveraged across 2M+ websites and applications globally, processing over one trillion events per month.

More than 250 companies, including Experian, Auto Trader, Strava, Condé Nast, and Hello Fresh, rely on Snowplow to build a well-governed, first-party data foundation that powers their customer-facing AI agents, in-session personalization and recommendations, and real-time analytics.

This Opportunity Snowplow is rebuilding how the company runs its own work. AI-native transformation is the operating priority for the business, and the Applied AI team owns it: the platform every team's AI work runs on, the pattern for how a function's critical workflows move onto agents, and the measurement of whether it worked. Individual AI adoption is already high across the company.

The next stage is redesigning functions so that agents run the recurring work and people set the goals and judge the output.

You are the first Forward Deployed Engineer on the team, reporting to the Senior Director of Applied AI. The team is also hiring a principal engineer to own the platform and an Applied AI engineer to embed with our Growth team. The Senior Director you report to still ships code. You are the one who moves between functions. You embed with one function at a time alongside its leader and the people who do the work by hand today, map how the work actually runs, build the agentic workflow that takes it on, tune it with the experts until they trust it, and hand it back to the function to own.

Then you move to the next one.

You sit with our major business functions, one at a time. The Senior Director of Applied AI and the function's leader choose each engagement, and your workflow maps are the evidence they choose from. Priorities will shift as the transformation moves, and the role is designed for that.

This is the forward deployed model applied inside our own company. You write the code, and you also do the discovery a consultant would. The function's leader owns the decision to change how their team works. You own the engineering and the pattern. You'll spend as much time listening to someone walk through a task they do by hand every week as you will writing the agents that run it, and the deliverable is a function that works differently, understands why, and can run what you built.

What you will ownThe workflow map. You sit with the people who do the work and leave with how it actually runs today: where work enters, what moves between steps, who says yes, and what "good" means. It is the first thing you do in every engagement.

The case for whether to build. Before anything gets built, you bring the evidence from the map and make the case for whether the workflow is worth moving onto agents, including when the answer is that it does not need AI. The decision sits with the function's leader and the Senior Director; the evidence is yours.

The build. You design and build agentic workflows on Snowplow's shared factory framework, which Applied AI runs: declared queues and gates, focused agents, and a central ledger that records every agent run. Every factory you build is compatible with every other one, in any function.

Approval gates. You decide where the agent acts on its own, where it proposes and a person approves, and what gets recorded so anyone can see what the agent did and why. The people in the function keep the judgment. The agents take the rest.

Agent readiness. You make the function's knowledge, systems of record, and quality bar explicit enough for agents to work from. Often this is the real work, and the agents are the easy part.

The last twenty percent. Most of the work comes after the prototype: reliability, edge cases, evaluation, and a team choosing it over the way they worked before. You tune the output alongside the domain experts and encode their judgment into checks that hold after they step back.

The handoff. You train the function to run and evolve what you built, then leave. Where an Applied AI engineer is embedded in the function, you hand it to them. Success is a workflow still running months after you have moved on, owned by people who are not you. After the handoff you stay available: when the function sees a problem or a way to improve the workflow, you coordinate it with the embedded and platform engineers…