Track
◈ Applied AI / Forward Deployed Engineer
The flagship track. Built on the real 2026 interview bar at OpenAI, Anthropic, Palantir and top AI startups — LLM foundations, RAG, agents, evals, production, system design, and the forward-deployed craft.
Readiness0%
Lessons0/40
Cards mastered0/70
Focused study263 min
Roadmap
11 modules1
The Role
2
LLM Core
3
Prompting
4
RAG
5
Agents
6
Evals
7
Production
8
Fine-tune
9
Sys Design
10
FDE Craft
11
The Rounds
Modules
clear 60% to unlock the next◈MODULE 01
The Role: Applied AI & Forward Deployed Engineer
What the job actually is, how it differs from ML/research, and how the interview loop works.
0%
🧠MODULE 02 · 🔒
LLM Foundations
Tokens, embeddings, context windows, sampling, training stages — the mental model you must have.
🔒 clear more of the previous module
✎MODULE 03 · 🔒
Prompt Engineering
Turn vague asks into reliable model behavior: structure, examples, chain-of-thought, structured output.
🔒 clear more of the previous module
❈MODULE 04 · 🔒
Retrieval-Augmented Generation
Ground models in real, private, current data: chunking, embeddings, vector search, reranking, evaluation.
🔒 clear more of the previous module
⛓MODULE 05 · 🔒
Agents & Tool Use
Give models the ability to act: function calling, agent loops, planning, multi-agent, MCP, and failure control.
🔒 clear more of the previous module
⚖MODULE 06 · 🔒
Evaluation & Quality
The skill that separates hobbyists from engineers: measuring LLM quality rigorously.
🔒 clear more of the previous module
⚙MODULE 07 · 🔒
Production: Latency, Cost, Reliability & Safety
Everything between 'cool demo' and 'runs for 100k users': cost, speed, reliability, security.
🔒 clear more of the previous module
🔧MODULE 08 · 🔒
Fine-Tuning & Model Adaptation
When prompting/RAG isn't enough: SFT, LoRA/PEFT, DPO, distillation — and when NOT to fine-tune.
🔒 clear more of the previous module
▦MODULE 09 · 🔒
System Design for AI Applications
The whiteboard round: design a production AI system end to end, with tradeoffs.
🔒 clear more of the previous module
🤝MODULE 10 · 🔒
Forward-Deployed Craft & Behavioral
The human half of the job: scoping ambiguity, customer communication, and nailing the behavioral round.
🔒 clear more of the previous module
🎯MODULE 11 · 🔒
Round-by-Round Playbook
The four rounds that decide 2026 loops: decomposition, take-homes, AI-assisted & debugging rounds, and the discovery call.
🔒 clear more of the previous module