The field guide for people who deploy AI at customers.

type Job = {
  company: "Anthropic";
  title: "Forward Deployed Engineer";
  location: "Paris, France";
}

type Job = {
  company: "Anthropic";
  title: "Forward Deployed Engineer";
  location: "Munich, Germany";
}

type Job = {
  company: "Anthropic";
  title: "Forward Deployed Engineer";
  location: "London, UK";
}

type Job = {
  company: "Anthropic";
  title: "Manager, Forward Deployed Engineering";
  location: "London, UK";
}

type Job = {
  company: "Scale AI";
  title: "Deployment Strategist - DHS/DOJ";
  location: "Washington, DC";
}

type Job = {
  company: "Scale AI";
  title: "Deployment Strategist, Space Systems Command";
  location: "Los Angeles, CA";
}

type Workflow = {
  category: "Discovery";
  steps: 5;
}

1. Record the call with an approved notetaker (e.g. Fathom, Granola).
2. Paste the transcript into your AI tool with the 'deal summary' prompt.
3. Ask it to list every assumption not stated by the customer.
4. Fix gaps, then generate a 5-bullet brief for your AE and a follow-up email draft.
5. Log the brief in CRM and set next-step dates.

type Workflow = {
  category: "Discovery";
  steps: 5;
}

1. Collect the company's site, latest 10-K or press releases and the attendees' LinkedIn roles.
2. Prompt: summarize business model, likely priorities, recent changes.
3. Generate 10 discovery questions ranked by importance.
4. Flag 3 'not a fit' signals to watch for.
5. Save it as a one-page call plan.

type Workflow = {
  category: "Demo prep";
  steps: 5;
}

1. List the top 3 pains from discovery.
2. Prompt the demo-story template to map pain → feature → outcome.
3. Generate fake but realistic demo data as CSV and load it into your sandbox.
4. Run the 'pre-mortem my demo' prompt as the skeptical buyer.
5. Record a 3-minute Loom of the flow for async stakeholders.

type Workflow = {
  category: "POC";
  steps: 5;
}

1. Draft the scope with the POC prompt (objective, criteria, timeline, owners).
2. Pressure-test each success criterion for measurability.
3. Review with the champion; agree on pass/fail before kickoff.
4. Turn the timeline into a shared checklist with dates.
5. Send a recap email confirming criteria in writing.

type Workflow = {
  category: "POC";
  steps: 5;
}

1. Pick one narrow task and collect 20–50 real example inputs (with permission).
2. Build the first version with an agent framework (e.g. LangGraph, OpenAI Agents SDK).
3. Write expected outputs and run them through an eval tool (e.g. promptfoo, DeepEval).
4. Trace failures with an observability tool (e.g. Langfuse, LangSmith).
5. Report pass rate against the agreed criteria, not vibes.

type Workflow = {
  category: "Security";
  steps: 5;
}

1. Gather your approved security docs (SOC 2 summary, DPA, architecture).
2. Prompt the AI to answer only from those docs and mark gaps NEEDS REVIEW.
3. Route NEEDS REVIEW items to security/legal owners.
4. Store final answers in your RFP tool (e.g. Loopio, Responsive) for reuse.
5. Track turnaround time per questionnaire.

type Prompt = {
  category: "Discovery";
  vars: ["COMPANY", "INDUSTRY", "PASTE", "ROLE OF ATTENDEES"];
}

I have a discovery call with [COMPANY], a [INDUSTRY] company, with [ROLE OF ATTENDEES]. Our product does [ONE-LINE PRODUCT]. Using only public info I paste below, list: 5 likely business pains, 10 open-ended discovery questions ranked by importance, and 3 signals that this is NOT a fit. Info: [PASTE]

type Prompt = {
  category: "Discovery";
  vars: ["PASTE"];
}

Here are raw notes from a discovery call: [PASTE]. Rewrite them as: current state, desired outcome, metrics they care about, decision process, technical environment, risks, and open questions. Flag anything I assumed but they did not say.

type Prompt = {
  category: "Discovery";
  vars: ["PASTE"];
}

From this email thread and call notes [PASTE], build a stakeholder map: name, title, what they care about, likely objection, influence (high/med/low). Suggest one question to ask each person next.

type Prompt = {
  category: "Demo prep";
  vars: ["LIST"];
}

Our product features: [LIST]. Prospect pains from discovery: [LIST]. Build a 20-minute demo flow: opening hook tied to their top pain, 3 scenes (pain → feature → outcome), and a close asking for a specific next step. Cut any feature that doesn't map to a pain.

type Prompt = {
  category: "Demo prep";
  vars: ["COLUMNS", "INDUSTRY", "USE CASE"];
}

Generate 25 rows of realistic but fake sample data for a [INDUSTRY] demo of [USE CASE] as CSV with columns [COLUMNS]. Include 3 edge cases that would make the demo more convincing. No real company or person names.

type Prompt = {
  category: "Demo prep";
  vars: ["BUYER ROLE", "PASTE"];
}

Here is my demo outline: [PASTE]. Act as a skeptical [BUYER ROLE]. List the 7 moments where you would lose interest or push back, and how I should fix each.

type Skill = {
  skill: "Command line, git, debugging basics";
  resource: "MIT Missing Semester";
  cost: "free";
}

type Skill = {
  skill: "Git branching, interactively";
  resource: "Learn Git Branching";
  cost: "free";
}

type Skill = {
  skill: "CS fundamentals";
  resource: "Harvard CS50x";
  cost: "free";
}

type Skill = {
  skill: "Python";
  resource: "LearnPython.org";
  cost: "free";
}

type Skill = {
  skill: "SQL";
  resource: "SQLBolt";
  cost: "free";
}

type Skill = {
  skill: "HTTP & APIs";
  resource: "MDN HTTP docs";
  cost: "free";
}

type Tool = {
  name: "Navattic";
  category: "Interactive demos";
}

type Tool = {
  name: "Consensus";
  category: "Demo automation";
}

type Tool = {
  name: "Reprise";
  category: "Demo sandbox/overlay";
}

type Tool = {
  name: "Walnut";
  category: "Interactive demos";
}

type Tool = {
  name: "Storylane";
  category: "Interactive demos";
}

type Tool = {
  name: "Supademo";
  category: "Interactive demos";
}

type Article = {
  source: "The Pragmatic Engineer";
  type: "Article";
}

type Article = {
  source: "SVPG";
  type: "Article";
}

type Article = {
  source: "a16z";
  type: "Article";
}

type Article = {
  source: "a16z";
  type: "Article";
}

type Article = {
  source: "Ramp Builders";
  type: "Article";
}

type Article = {
  source: "Szymon Paluch";
  type: "Article";
}

type Repo = {
  repo: "modelcontextprotocol/servers";
  stars: 91313;
}

type Repo = {
  repo: "openai/openai-cookbook";
  stars: 76649;
}

type Repo = {
  repo: "microsoft/autogen";
  stars: 61443;
}

type Repo = {
  repo: "crewAIInc/crewAI";
  stars: 59768;
}

type Repo = {
  repo: "anthropics/claude-cookbooks";
  stars: 53495;
}

type Repo = {
  repo: "run-llama/llama_index";
  stars: 52524;
}

type Post = {
  handle: "@lakhera2015";
  date: "2026-09-17";
}

🚀 Forward Deployed Engineer might be one of the most interesting roles to watch in the AI job market right now.

A few years ago, when we talked about Forward Deployed Engineers, one company immediately came to mind: Palantir. That is changing very quickly.

I spent some time

type Post = {
  handle: "@DataScienceDojo";
  date: "2026-10-08";
}

We're excited to have Aater Suleman, Co-founder of Vixul, Hannah Marlowe, Senior Director of AI Forward Deployed Engineering at Databricks, and Dan Maloney, CEO of LandingAI, joining the panel "Secure AI Needs Forward Deployed Engineers" at the Agentic AI Conference! 🎙️

Secure,

type Post = {
  handle: "@rocrizzardini";
  date: "2026-10-10";
}

You hired a forward deployed engineer to write code. The survey says the job is 31% code. The other 47% is why the role exists.

Perspective AI surveyed 1,500 forward deployed engineers at 154 companies, February to April 2026 (published May 18, 2026):

1. The median week: 47%

type Post = {
  handle: "@mardehaym";
  date: "2026-10-07";
}

Hot take: forward-deployed engineering is overrated.

Gartner predicts that by 2028, 70% of enterprises will abandon agentic AI built by vendor forward-deployed engineers, citing rising costs and the inability to evolve those systems independently.

I’ve spent over a decade

type Post = {
  handle: "@AlfinCodes";
  date: "2026-10-11";
}

Anthropic is hiring Forward Deployed Engineers (FDEs)

But what the hell is a Forward Deployed Engineer?

type Post = {
  handle: "@zahiruddin_t";
  date: "2026-10-12";
}

Everyone talks about Forward Deployed Engineers at OpenAI, Anthropic, and Palantir.
But I think there's an equally important role 
The Internal Forward Deployed Engineer.
The engineer who helps teams across an enterprise adopt AI.

type Pay = {
  metric: "Median advertised FDE salary (US)";
  value: ">$188,000";
}

type Pay = {
  metric: "FDE posting growth, Jan–Aug 2026 vs 2025";
  value: ">1,000%";
}

type Pay = {
  metric: "Indeed FDE postings Apr 2025 → Apr 2026";
  value: "+729%";
}

type Pay = {
  metric: "US FDE postings, first 7 months of 2026";
  value: "5,200+";
}

type Pay = {
  metric: "Companies planning FDE hires";
  value: "70%";
}

type Tip = {
  title: "Give the model the role and the reader";
}

Say who it is writing as and for whom (e.g. 'a CFO who skims'). Output changes more from audience than from adjectives.

type Tip = {
  title: "Paste the source, don't describe it";
}

Paste call notes, docs or logs. Summaries of summaries lose the details that matter.

type Tip = {
  title: "Ask it to flag gaps instead of guessing";
}

Add 'If the input doesn't support an answer, write NEEDS REVIEW.' This cuts made-up details in customer-facing work.

type Tip = {
  title: "Use structure for reusable outputs";
}

Request fixed sections or a table so status updates and POC docs look the same every week.

type Tip = {
  title: "Iterate in two passes";
}

First ask for a draft, then ask it to critique the draft as the skeptical buyer, then revise.

type Tip = {
  title: "Keep a personal prompt library";
}

Save prompts that worked with your [BRACKET] variables. Our prompts page is a starting point.