0% completed
The FDE Teams and What Each One Ships
On This Page
One Job, Several Teams
GenAI
Data Cloud
Applied AI
Developer AI
Generative Media and Other FDE Roles
Comparing the Teams
Using This in an Interview
Key Takeaways
Practice Questions
Search Google's careers site for forward deployed engineer roles, and the titles do not match each other.
| Posting title |
|---|
| Forward Deployed Engineer III, Generative AI, Google Cloud |
| Forward Deployed Engineer IV, Data Cloud, Google Cloud |
| Forward Deployed Engineer IV, Applied AI, Google Cloud |
| Staff Forward Deployed Engineer, Developer AI, Google Cloud |
| Staff Forward Deployed Engineer, Generative Media, Google Cloud |
The level appears after the job name, and the team appears after the comma. The team decides what you will build for customers and which skills the posting requires. It also decides which kind of system you should be ready to design in the interview.
One Job, Several Teams
Every Google Cloud FDE team hires an embedded builder who takes a customer from prototype to production. Each team builds on a different part of Google's products. Most postings share their wording, including the "white glove" deployment and the feedback loop to Google's product road map. The differences sit in the responsibilities and the required experience. One prep guide reports that the interview covers the same areas for every team. The questions and emphasis change with the team.
The freight carrier from What a Forward Deployed Engineer Does at Google Cloud shows the difference. Four of these teams could each run a separate engagement at that one customer.
GenAI
The GenAI team ships agentic systems: agents and sub-agents, the MCP servers they use, and the evaluation and observability that keep them reliable. Its postings describe the FDE as the lead developer who moves rapid prototypes to production-grade agentic workflows. They also name the "connective tissue" to the customer's APIs, older data stores, and security perimeters. Preferred experience includes multi-agent frameworks such as Google's Agent Development Kit (ADK) and LangGraph. The Staff posting adds secure tool calling with MCP and OAuth, the standard way to let a user grant a program limited access.
At the carrier, this is the exceptions agent from the first lesson.
Data Cloud
The Data Cloud team ships data and machine learning workloads, and the data foundation that agents depend on. Its postings require Python and SQL, plus experience with Spark, BigQuery, Spanner, or AlloyDB. The work includes three kinds of task:
- Migrations. Moving expensive analytics or database workloads onto Google Cloud, with the cost and speed measured before and after.
- A data foundation for agents. Building metadata, descriptions of what each table means, and access rules, so an agent reads only trusted data it is allowed to see.
- Machine learning models. Training and evaluating them on customer data, with monitoring in place.
The postings also ask the FDE to present working results on the customer's own data after removing sensitive fields such as payment card and health records. At the carrier, this team would move the billing warehouse to BigQuery and prepare it so an agent can answer billing questions safely.
Applied AI
The Applied AI team ships conversational AI and customer experience systems: agents that talk to a company's own customers. Its FDE IV posting describes taking conversational prototypes to production and defining the first customer user journeys (CUJs) at the customer's site. A customer user journey is one complete path a user follows, such as "a shipper asks where a delivery is and receives an answer".
This team's posting is the most operational. It requires Terraform, a tool that creates cloud resources from configuration files. It also asks for experience tracing a conversation id across services to fix failures as they happen. It also lists troubleshooting live systems with heavy traffic and up to 30 percent travel. At the carrier, this team would build an agent that answers shippers directly, so fewer questions reach the desk at all.
Developer AI
The Developer AI team ships agentic workflows for a customer's own software engineers. Its Staff posting names Gemini Code Assist, Google's Antigravity development environment, and the SDK and command-line tools beneath them. The work includes pipelines that turn a written specification into a pull request, large refactors, language migrations, and automated loops from an incident report to a fix. These connect to tools the customer already uses, such as GitLab, ServiceNow, and Slack. The connections use MCP and A2A, a protocol that lets agents from different systems work with each other.
The FDE first measures what slows the engineering organization down, such as review delays or low test coverage. Then the FDE agrees on a success metric with the customer. At the carrier, this team would work with the carrier's 40 software developers, for example on moving the old TMS integration code to a supported language.
Generative Media and Other FDE Roles
Some FDE roles sit outside the main Cloud teams or serve a narrower kind of customer. The Generative Media Staff posting asks for 3 years of building with or fine-tuning models that generate images, video, or audio. A freight carrier would rarely need this team, while a retailer that produces thousands of product images would.
Google also posts FDE roles in Google Labs, YouTube, Google Public Sector, and a fixed-term Kaggle and DeepMind role. Their requirements differ from the Cloud teams. The YouTube GenAI posting, for example, asks for full-stack development and front-end languages such as JavaScript or TypeScript.
Comparing the Teams
| Team | What it ships | A requirement the others lack |
|---|---|---|
| GenAI | Agents, MCP servers, evaluation | Multi-agent frameworks, OAuth for tools |
| Data Cloud | Data and ML workloads, migrations | SQL, Spark, measured price-performance |
| Applied AI | Conversational and customer experience agents | Terraform, live debugging, travel |
| Developer AI | Agentic workflows for software teams | SDLC metrics, A2A, developer tools |
| Generative Media | Image, video, and audio generation | Fine-tuning generative models |
Using This in an Interview
Recruiters and interviewers often ask, "Why this team?" Answer by naming the team's work and connecting it to something you have built.
"I applied to the Data Cloud team because most AI projects I have seen fail on the data, not the model. In my last role I moved a 9-terabyte reporting workload to a new warehouse, measured cost and run time before and after, and then built the table descriptions an analytics agent needed. That is the work the posting describes."
Expect two follow-up questions:
- "Would you consider the GenAI team instead?" Say yes if you can, and name the overlap: both teams build data foundations and evaluation, but GenAI owns the agent itself.
- "What would you build first at a new customer?" Name the team's first deliverable: for Data Cloud, an inventory of the data the agent will need and who may see it.
Key Takeaways
- Google posts FDE roles by team, and the title names it: GenAI, Data Cloud, Applied AI, Developer AI, or Generative Media.
- GenAI ships agents, Data Cloud ships data and ML foundations, Applied AI ships customer-facing conversational agents, and Developer AI ships workflows for software engineers.
- One customer can need several teams, as the freight carrier shows.
- Read the team in the title before you prepare, because it decides which systems you should be ready to design.
The next lesson, Forward Deployed Engineers, Customer Engineers, and Other Google Cloud Roles, compares the FDE with the Google Cloud roles that meet the same customers earlier.
Practice Questions
1. A posting asks for Terraform, tracing conversation ids, and up to 30 percent travel. Which team is it most likely from, and why?
<details><summary>Show answer</summary>Applied AI. Its posting is the most operational: it builds conversational agents at the customer's site and fixes live failures, which needs infrastructure as code and tracing across services.
</details>2. A bank wants an agent that answers questions over its transaction tables, but no one knows which tables are trustworthy. Which team should start, and with what?
<details><summary>Show answer</summary>Data Cloud, with the data foundation. The agent needs table descriptions and access rules before it can answer safely, and that work is what Data Cloud postings describe.
</details>3. Why can the same customer need both a GenAI engagement and an Applied AI engagement?
<details><summary>Show answer</summary>Because they serve different users. At the carrier, the GenAI agent helps the internal desk resolve exceptions, while the Applied AI agent answers shippers directly so fewer questions reach the desk.
</details>Reading Progress
0%
On This Page
One Job, Several Teams
GenAI
Data Cloud
Applied AI
Developer AI
Generative Media and Other FDE Roles
Comparing the Teams
Using This in an Interview
Key Takeaways
Practice Questions