Door 01 · Learn

The FDE craft, made learnable.

The instincts Palantir trained with a plane ticket — turned into something you can study. Not a generic bootcamp: a curriculum that compiles from a real job description, proves mastery with evals, and leaves you with a portfolio instead of a certificate.

The competency map

Six clusters, one ladder.

Every Forward Deployed Engineer role decomposes into these. Each cluster ships both halves — the human knowledge you internalize and the agent tool you build as you learn (which becomes the portfolio artifact that proves it).

01

Production SWE

Ship real systems in Python and TypeScript — tests, types, CI, observability. The bar is production, not toy.

02

Enterprise / hybrid deploy

Docker, Kubernetes, CI/CD across public cloud, VPC, and on-prem. The perimeter constraints that define FDE work.

03

Modern AI stack

RAG, vector DBs, evaluations, fine-tuning. Evaluation is the load-bearing skill — “how do we know it works at 5pm?”

04

Agentic system design

Agent loops, tool-use, MCP, orchestration, guardrails — integrated deeply with enterprise infrastructure.

05

The FDE craft

Discovery, reading the political terrain, finding the data reality, scoping the smallest end-to-end win. High agency.

06

Leadership & playbooks

(For lead roles.) Define standards, mentor, and codify evolving playbooks that scale the practice.

Validated against real roles

A JD is the input.

The course doesn’t guess what matters — it’s judged by one question: would someone who finishes this be ready to land this role? Drop a job description, get a target + a prep curriculum + the tools per competency.

Prove it

Eval-as-gate, not a certificate.

You don’t “finish” the course — you pass its gate. Self-score each cluster against the JD; readiness is a number, and the proof is a deployed artifact plus a forkable agent tool.

The by-product that gets you hired: a Site Survey, an agent architecture, a deployed service ×2, a RAG system + eval scoreboard, a working multi-tool agent + MCP server — plus the agent tool you built for each. That toolkit is the interview.