Skip to content
Forward Deployed

Forward Deployed

Start

Choose Your Path

A path marks each module as required, recommended, optional, or skip for your background, so you know what to study first. It is optional: every lesson is open without one, and you can change it any time.

Interview this week? Go straight to the final-week review.

Take the Diagnostic

Ten questions about your background, your timeline and how you study. It takes about three minutes, and your answers stay on this page.

Or Pick Your Starting Point

Each card opens the path for that background, with every module flagged.

  • Beginner

    Some programming, little production or AI experience.

    Engineering foundations, then LLM systems step by step.

    Beginner path · about 118 h

  • Experienced software engineer moving into AI

    Strong backend, little ML or LLM experience.

    LLM systems, agents and the design method, without the basics.

    Experienced Software Engineer path · about 164 h

  • Data engineer

    Pipelines, SQL, Spark and governance.

    Retrieval, agents and evaluation; ingestion basics you already know.

    Data Engineer path · about 147 h

  • ML engineer going to production

    Models and training, weaker on systems and customers.

    Agents, production operations, design and customer framing.

    ML Engineer path · about 155 h

  • Architect or consultant

    Customer-facing, broad architecture experience.

    Hands-on AI depth, delivery and a portfolio as proof.

    Architect or Consultant path · about 169 h

  • Forward Deployed Engineer candidate

    Mixed background, interview in two to eight weeks.

    The full core: systems, design, delivery, stories and mocks.

    Forward Deployed Engineer path · about 169 h

  • Revision-only candidate

    Already studied; interview this week.

    The opening routine, the design method and final-week drills.

    Revision Only path · about 41 h

  • Forward Deployed Engineer: Frontier Lab Loop

    The Forward Deployed Engineer path for loops with a coding round and a take-home build of 3 to 5 hours, with customer delivery and evaluation required.

    Loop variant of the Forward Deployed Engineer path · about 199 h

How the Site Works

Four areas, each with one job. A path tells you which modules to start with; the areas are where you study them.

  • Learn

    The curriculum, by track and module. Read a lesson, then rate how confident you feel so the revision queue knows what to bring back.

  • Practice

    Worksheets you fill in, then compare with a model answer and score honestly. The behavioural questions are answered with your own stories.

  • Design

    Complete system designs. Read one after practising the same kind of problem, or when a worksheet stumps you.

  • Interview

    Timed mock rounds with a scorecard, topic hubs, and the final-week review for the days before the loop.

The Practice Loop

  1. Read the prompt and resist the urge to design

    The prompt is vague on purpose, like a real customer conversation. Ask what breaks before you reach for a vector database.

  2. Fill in the blanks before you reveal anything

    Writing a wrong answer and seeing why it is wrong teaches more than reading a right one. Type fast; an empty worksheet teaches nothing.

  3. Reveal one section at a time

    Open each model answer only after you have attempted that section. It shows the ground a strong answer covers, not the only correct answer.

  4. Score yourself against the rubric, honestly

    Most first attempts score a 3. The row that stays lowest across several scenarios is the one to work on.

  5. Say the closing answer out loud

    Every model answer ends with a two-minute spoken summary. Say yours before you reveal it: being right but unable to say it in two minutes reads as not knowing it.

What the Interview Tests

The same five things are scored in every round. They are the rows of the scorecard.

Problem framing
Do you find the dangerous constraint before you design, or jump to tools?
Architecture
Are your components specific, bounded and observable, or is it a generic diagram?
Evaluation
Can you say how you would know it works, with datasets, thresholds and owners?
Production thinking
Do you plan for failure, rollback, cost and drills, or does the story end at the demo?
Communication
Can an executive and an engineer both take what they need from the same answer?