Skip to content
NextiaNextia Learning

AI System Design and Cost Engineering

Design an AI product before you build it, and know what it will cost. You write Larkfield's design and cost proposal for a document assistant that several shops share, one module at a time. Every number in it is measured, a published price or a stated assumption.

Start course See the curriculum

Read every lesson as a guest, or sign in to enroll and track your progress.

What you'll learn

By the end of this course, you will be able to:

  • Translate business requirements into a coherent architecture.
  • Estimate capacity and cost with explicit assumptions.
  • Compare model, retrieval and infrastructure options.
  • Defend design decisions and their limitations.

About this course

Larkfield’s help desk has an assistant that answers questions from the shop’s own documents. Now Bramble Books wants the same thing, and Omar wants a third and a fourth shop next year. Before anyone builds more, Omar asks Kwame two questions: “What exactly will we build?” and “What will it cost per shop?”

In this course, you write Kwame’s answer: a design and cost proposal for a document assistant that several shops share. First you write down what the users need and turn vague wishes into targets you can measure. Then you draw the architecture and follow each shop’s data and identity through it.

Next you put numbers on the design. A small calculator turns demand into concurrency, storage and tokens. Then it turns them into a cost table, the cost per question and the cost per shop, for a low, a base and a high case. You compare cheaper and faster options with measured results, not guesses.

Finally you plan for failures and for growth, record your decisions, and present the proposal to people who are not engineers. Every number has a label: measured (from a real run), price (from a published price list) or assumption (a stated input with a range). A design that states its assumptions can be checked; one that hides them cannot.

Before you start

You completed RAG, Databases and Storage for AI Applications, Scaling APIs and AI Workloads, and Authentication, Permissions and Multi-Tenant Applications, or you have the same skills. You do not need an account or money. The cost calculator is a small Python program with no extra packages, and every AI measurement comes from real calls that the course team recorded.

Courses to take first: RAG: Building AI That Uses Your Data, Databases and Storage for AI Applications, Scaling APIs and AI Workloads, Authentication, Permissions and Multi-Tenant Applications.

Who this is for

People who have built a RAG assistant, a database-backed API, a queue and a multi-tenant sign-in, as in RAG, Databases and Storage for AI Applications, Scaling APIs and AI Workloads, and Authentication, Permissions and Multi-Tenant Applications. Now they must decide what to build next, what it costs and what can go wrong. Developers, AI engineers and solution architects who write or review design proposals.

Lead or use AI, but do not build it?

You can study the ideas of this course without writing code. Read modules 1–3 and 6 (about 7–9 hours), answer their knowledge checks, and get a free concepts certificate. You can come back for the practice later.

The paths for people who lead or use AI take this course this way. See the paths

How we made this course

Experienced AI practitioners designed this course. AI helped us draft the content; people chose the teaching approach and the examples, and reviewed every lesson. The outputs that you see are the actual results of running the code. How we make our courses

Inside the course

Module 2ConceptsDesign the end-to-end architecture

3 lessons

Draw the system. Give every part one job, trace each tenant's data and identity through requests and background work, and decide what stays in one application until a measurement says otherwise.

  1. Service boundaries55 min
  2. Data and identity flows55 min
  3. Simplicity and evolution50 min

Module 3ConceptsEstimate resources and cost

3 lessons

Put numbers on the design. Turn traffic into concurrency, storage and tokens, build a cost model by driver from list prices and measurements, and find the cost per question, per document and per tenant.

  1. Workload arithmetic55 min
  2. Cost model60 min
  3. Unit economics55 min

Module 4Optimize with measured trade-offs

3 lessons

Make it cheaper or faster, with proof. Route questions between a small and a strong model, cache and batch work, and compare managed services with running things yourself, including the people's time.

  1. Model routing60 min
  2. Caching and batching55 min
  3. Build versus buy55 min

Module 5Design for failure and growth

3 lessons

Plan for bad days and for growth. Walk through six failures with owners and mitigations, set recovery targets and test a restore, and grow in steps that measurements trigger.

  1. Failure scenarios55 min
  2. Recovery objectives50 min
  3. Growth plan50 min

Final module · choose at least oneCase studies

2 case studies

Two end-to-end case studies on open data. Plan cost and capacity from a real public demand series, and review the design of a document-question service for a real open document collection. Complete at least one for the certificate; each one is a portfolio piece.

Complete at least one case study for the certificate. The others are optional. What each one covers

  1. Cost and capacity plan from real demand data120 min
  2. Design review of an open document-question service180 min

After the case studiesFinal assignment

Use every module in one practice project. You collect evidence, and check it against the course outcomes.

  1. Final assignment180 min

Case studies

Realistic projects from start to finish, on real data, with every output explained. Each one makes a portfolio piece. Complete at least one for the certificate; the others are optional.

  1. Cost and capacity plan from real demand data2 hr

    Turn a public demand series into peaks, growth and a cost range for a public question-answering assistant, with the course's calculator.

  2. Design review of an open document-question service3 hr

    Size storage and ingestion for a real open document collection, choose routing and caching with measured numbers, and write the decision records and the failure table.

Reviews

Loading the reviews…

    Took this course? Sign in to review it. Learners can write a review after they complete 25% of the lessons.

    Share this course