AI
companies are paying
$1 Million for this role
BEST
Paying Jobs
in Tech
FORWARD DEPLOYED
ENGINEER
called
the exact
eight week
roadmap
to get in.
save this reel.
Week-1
Data engineering
Select Star SQL
This is an interactive book which aims to be the best place on
the internet for learning SQL. It is free of charge, no sign up,
and doesn't require registration or downloads. It helps you learn by
running queries against a real-world dataset to complete
projects of consequence. It is not a mere reference page - it
conveys a mental model for writing SQL.
I expect little or no coding knowledge. Each chapter is designed
to take 30 minutes. As most of the world's data is stored in
databases, this time will pay rich dividends!
This is the exact
course
Week-2
Cloud & infra
docker.
Terraform
kubernetes
that
And this is
free on
Google Skills
Skill up on
AI today
AI isn't a future state. It's here to stay. With more than a
decade of AI innovation and integration in the cloud,
Google is your trusted partner in navigating this new
reality.
featured
Artificial Intelligence
Vertex AI
Gen
Cloud
Boost
Week-3
Build with the Claude API
ANTHROP\C
Building with the Claude API
This comprehensive course covers the
full spectrum of working with Anthropic
models using the Claude API
build with
Claude API
Week-4
Production RAG &
vector databases
Phonometrics
Learn to love vectors
What is Embeding Engineering?
AI
RAG from Scratch
Advanced Techniques
Building and Evaluating Advanced RAG
focus on
production RAG
vector databases.
the links.
Week-5
Agents, MCP
ANTHROP\C
Hugging Face
and MCP.
Anthropics Claude
the best
you can get
Week-6
Evals and observability
All Courses Short Course Beginner
Assess AI
Agent
Performance
Evaluating AI Agents
Instructors: John Dilullo, Aman Khan
arize
AI
Engineering at Anthropic
Building effective agents
We've worked with dozens of teams building
LLM agents across industries. Consistently,
the most successful implementations use
reusable patterns rather than
frameworks.
Week-7
Ship it in a
messy environment
kaggle
FastAPI
Identify
from Kaggle
deploy it
and FastAPI
repo.
Awesome Forward
Employment Engineering (FDE)
Forward Employment Engineering is the art of shipping complex
AI models into production. Software
Architect, or Technical Special Ops,
who bridge the gap. (The
technical core product and a client's messy, real-world
AI Action Letter #45: An Engineering-First
Plan for the Fastest-Growing AI Job, Built
Backward From Real Palantir, OpenAI, and
Anthropic JDs
ABHIJAY ARORA VUYURRU
JUL 4, 2026 AT 6:57 PM
Listen
Hey folks,
A few weeks back I broke down the Forward
Deployed Engineer role. What it is, why it
blew up, and what it pays. That letter got a
huge response. And almost every single
reply said the same thing: "Okay Abhijay, but
HOW do I actually become one?"
So here's the answer. The full
engineering plan. Eight weeks, almost entirely free.
Quick refresher in case you missed the first
one. FDE postings jumped more than
1,000% last year. Palantir starts around
$135K to $200K, and at Anthropic total comp
runs $300K all the way to $1.2 million.
One thing I need to say up front, because I
get asked it constantly. This is not the AI
Product Manager path. An FDE ships
production code inside a customer's messy
systems. So this plan is engineering-first.
Real code. Real data. Real deployment. If you
want the product route, that's a different
letter.
But before you study anything, one warning
from the field. I went digging through the
Reddit threads where actual FDEs talk, and
the top comments are blunt. One calls it
"sales with extra steps." Another calls it "a
consultant role Palantir made sound cool."
So the truth sits in the middle. Some FDE
jobs are incredible, but all are rebranded
sales quotas with a nicer title. Your very first
skill is telling them apart.
Then I did the boring, useful thing. I pulled
the actual job descriptions from Palantir,
OpenAI, and Anthropic and built this plan
backward from them.
First: filter the fakes
Before you study anything, learn to read the
job description. Skip a posting if it has:
• A quota, OTE, or commission structure
(that's a sales job wearing a costume)
• Reporting into Sales or GTM instead of
Engineering
• Responsibilities that are mostly demos
and closing deals
A real FDE builds. They take a company's
product and make it actually work inside a
customer's messy systems.
There's a name for that gap. The space
between what a product does out of the box
and what the client actually needs to
succeed. Palantir calls it "the Delta." Your
whole job is closing the Delta, and it takes
three hats: part software engineer, part data
architect, part consultant. This plan trains all
three.
What Palantir, OpenAI, and Anthropic
actually ask for (I read the JDs)
• Palantir FDSE: "strong coder" in Python,
Java, C++, or TypeScript. Owns projects
end to end like a startup CTO. Wranglés
massive-scale data, builds custom web
apps, monitors production. 1+ years
experience. Travel up to 25%.
• OpenAI FDE: owns discovery, scoping,
system design, build, and production
experience with LLMs including agent
development and evaluation
frameworks. Ships real artifacts like MCP
servers, sub-agents, and agent skills.
Python plus ideally TypeScript. 3+ years.
Three themes jump out. Deep engineering
(code, data, cloud). Agents and MCP. And
evals. This plan is built around exactly those.
One honest note: Palantir takes people with
1 year of experience, so it's the most
reachable door. OpenAI and Anthropic
usually want 3 to 5+. Aim where you are.
Part 1: The three engineering
hats (Weeks 1 to 6)
Week 1: Data engineering (the mission
starts here)
Most AI roadmaps skip this. Don't. An FDE
mission almost always starts with a data
audit, not a model. Palantir's JD literally says
you'll be "wrangling massive-scale data." If
you can't untangle a client's 20-year-old
data, you can't build a bridge with
free interactive book:
https://selectstarsql.com
• Data modeling for reality: star schema,
and the Bronze, Silver, Gold "medallion"
pattern for turning raw client data into
something usable.
• The modern data stack: get hands-on
with dbt (free "dbt Fundamentals"
course) and DuckDB for fast local
analysis of a client's messy CSVs.
• Anchor course: the free Google Cloud
Skills Boost "Data Engineer" path:
https://www.cloudskillsboost.google
Week 2: Cloud and infrastructure
(the vehicle)
FDEs get dropped into complex cloud
environments and have to deploy. This is
the week that separates you from someone who
only builds in a notebook.
Week-8
The FDE interview
Palantir
database.
Use
framework.
Watch mock interviews
DM hiring managers
loom video
your resume.
Don't worry
networking part.
I've put together
guide on
IDs of
managers,
post ideas
about
everything
and everything else.
Comment “FDE” to get the complete 8 week roadmap with links to all courses! #fde #Tech #usa