The Forward Deployed Engineer Track

Short, self-contained lessons, beginner→advanced at the FDE level. Each lesson teaches one idea tied to the mission — be genuinely ready for Forward Deployed Engineer (AI) work in three months — and every lab produces a real artifact (working code, an eval harness, a demo, a scoping doc) that goes into your portfolio repo and works as interview evidence. One continuous scenario threads the course: a fictional virtual-care provider whose systems you prototype, ground, evaluate, harden, and finally deliver as a full field engagement. Grounded in verified 2026 sources: real FDE job postings, current Anthropic and OpenAI engineering docs, and recognized practitioners on evals, agents, and production LLM systems.

Lessons

Module 1 · The FDE Role & Mindset · Beginner

• Lesson 01 What a Forward Deployed Engineer Actually Does Half engineer, half consultant, full owner — the role's Palantir origins, the 2026 market, the four competencies hiring managers actually test, and your first artifact: a gap map from three real job postings. ⏱ ~9 min lesson · 🎧 9 min listen · 🧪 ~45 min lab

Module 2 · Applied LLM Engineering · Intermediate

• Lesson 02 Prompting and Context Engineering System prompts as specs, structured outputs, the context window as a finite budget, context rot — and when prompting beats fine-tuning. ⏱ ~9 min lesson · 🎧 10 min listen · 🧪 ~60 min lab
• Lesson 03 Tool Use and Agents The workflow-vs-agent call, tool descriptions as an engineering discipline, MCP as the integration seam, and when multi-agent is the wrong answer — then build a tool-using triage agent and debug a deliberately bad tool. ⏱ ~10 min lesson · 🎧 10 min listen · 🧪 ~90 min lab
• Lesson 04 RAG and Grounding Retrieval quality over vector-db worship, grounding via mandatory citations and honest refusals — and knowing when RAG is the wrong tool. ⏱ ~9 min lesson · 🎧 11 min listen · 🧪 ~90 min lab
• Lesson 05 Evals That Prove It Works Why evals are the FDE's real deliverable — success criteria, the three-level framework, grader choice, validating an LLM judge, and proving your harness catches regressions. ⏱ ~10 min lesson · 🎧 9 min listen · 🧪 ~120 min lab

Module 3 · Rapid Prototyping & Integration · Intermediate

• Lesson 06 Zero to Demo The 0→1 craft — thin vertical slices, demo-driven scoping, honest demo theatre, and a clickable web demo shipped in days, not sprints. ⏱ ~9 min lesson · 🎧 10 min listen · 🧪 ~120 min lab
• Lesson 07 Integrating with Messy Enterprise Systems Legacy APIs, CSV exports of doom and auth mazes — the week-one data strategy, PII instincts, MCP servers as the integration seam, and data-quality triage. ⏱ ~10 min lesson · 🎧 11 min listen · 🧪 ~120 min lab

Module 4 · Production Hardening · Intermediate

• Lesson 08 Reliability, Guardrails, and Observability How LLM systems fail differently — prompt injection and the lethal trifecta, layered guardrails (input/output/human), and tracing with Langfuse — hardening a demo into production. ⏱ ~10 min lesson · 🎧 11 min listen · 🧪 ~90 min lab
• Lesson 09 Cost, Latency, and Eval-Driven Iteration Pilots die on unit economics — instrument cost per query and p50/p95, then cache, route, budget, and prune with your eval suite as the regression gate that makes fast iteration safe. ⏱ ~10 min lesson · 🎧 11 min listen · 🧪 ~90 min lab

Module 5 · Customer-Facing Craft · Advanced

• Lesson 10 Discovery, Scoping, and Demos That Land The consulting half of the role — discovery that finds the real problem behind the ask, the scoping doc, stakeholder translation, demos as meeting craft, and how to recommend against AI without losing the customer. ⏱ ~10 min lesson · 🎧 11 min listen · 🧪 ~90 min lab

Module 6 · The Move & Beyond · Advanced

• Lesson 11 Capstone: Field Engagement Run one full FDE engagement end to end — synthetic data, prototype assembly, eval evidence, hardening rerun, and the case study (with its required "not to build" section) that becomes your portfolio centerpiece. ⏱ ~9 min lesson · 🎧 10 min listen · 🧪 ~180 min lab
• Lesson 12 Landing the Role (Optional Track A) The FDE interview loops decoded — take-homes, the decomposition round, portfolio publishing, and an honest job-market map from Australia. ⏱ ~9 min lesson · 🎧 10 min listen · 🧪 ~90 min lab
• Lesson 13 Going Deeper (Optional Track B) The advanced roadmap after the course — when fine-tuning is actually warranted, multi-agent systems in production, staying current without drowning, and contributing patterns back. ⏱ ~9 min lesson · 🎧 10 min listen · 🧪 ~60 min lab

Reference

• Your portfolio Artifact Tracker Tick off the 13 lab artifacts as you produce them — the raw material of your portfolio repo and interview stories.
• Living document Glossary The canonical vocabulary for the course — the FDE role, context engineering, agents, RAG, evals, guardrails. Every lesson uses these terms exactly.