Free guide · 7-day email series

FDE 101

Forward Deployed Engineering for people who ship agent outcomes — when the role is the right motion, how to avoid a 55-repo dev shop, and how to keep multi-customer agent fleets isolated and alive.

1. What an FDE is

A Forward Deployed Engineer is a customer-facing software engineer — someone you would hire as an engineer and also trust in front of a customer.

The customer is not buying a SKU or a timesheet. They are buying an outcome. Implementation detail is your problem.

Pedagogy tracks Kevin Bai’s FDE 101 (Anthropic Applied AI; founding FDE at Rippling; ex-Palantir) — talk. This page is jurniti’s practical companion for agent deploy craft.

2. The 2×2 — when you need FDE

You need FDE in one awkward corner: a very technical product sold to non-technical buyers.

  • Technical product + technical buyers → DevRel / technical GTM
  • Configurable SaaS + non-technical buyers → sales-led
  • Technical platform + non-technical buyers → FDE

Ask need, not want. If you are not in that corner, do not stand up an FDE org because the title is trendy.

3. Design partnership at enterprise scale

Early startups find product-market fit with design partnerships. FDE is that idea scaled into the enterprise: sit with the customer, take context, ship something that works.

If every FDE builds from a blank repo per logo, you do not have an FDE function. You have a dev shop — 55 repos, maintenance death, engineers who quit. Different business. Call it what it is.

4. Platform primitives

FDEs assemble shared building blocks into outcomes. How atomic those primitives are depends on the industry — sometimes ~60% pre-built, sometimes extremely granular. The rule does not change: do not invent the floor every engagement.

For agent work, customer-scoped skills and memory can be bespoke. The runtime should not: isolation boundary, always-on compute, key custody, harness install, snapshot and fork.

5. The agentic shift (2026)

Nearly every platform is agentic. That means nearly every platform is customizable. That means customers often have no idea how to implement what you sold them.

Leaving success to their hands is how upmarket expansion stalls. FDE — or a brutal churn curve — sits in the gap between demo and production.

6. Deploy craft for agent fleets

Bespoke stays with the customer. Anything generalizable moves to the platform over time. Early FDE work is also product scouting.

Avoid:

  • Laptop fleets (battery, no isolation, can’t leave)
  • DIY VPS snowflakes (SSH is not a product)
  • Shared-kernel multi-tenant for production credentials

Prefer: hardware isolation per tenant, BYOK inside the runtime, always-on so agents survive when you leave the building, templates you can fork for the next logo.

7. The third way

Babysit a VPS forever — you become ops, not FDE.

Black-box host — someone else holds keys and proxies model spend.

Managed runtime, still yours — isolated microVMs, harness of your choice, keys in the tenant, templates and fork, flat monthly compute, 30-day money-back on first purchase. That is jurniti.

Prefer the paced version?

Seven emails. One idea a day. Jump-ahead links if you read fast.

Get Day 1