Applied AI infrastructure fundamentals
Deploy and operate one small AI service end to end, with a diagram, a cost estimate, an evaluation set, an alert threshold and a runbook another operator can follow.
Privileged access
8 modules
4 weeks
10 to 14 hours
What you will learn
- Explain how an AI request moves through application, model, data, storage and operations layers.
- Choose between hosted APIs, managed platforms and local open models using explicit tradeoffs.
- Estimate the compute, memory, storage, network and cost a small workload needs.
- Package, deploy, secure and observe a small service.
- Build a retrieval workflow whose answers cite their sources and whose failures are visible.
- Hand off a reproducible system with its limitations, its tests and an operator runbook.
Your instructor
Course content
Method step 1, Choose
Follow one request from the interface to the response, and name every part it passes through.
- Follow one request end to end
- Diagram a chatbot or document assistant
- The diagram kit and glossary
Method step 2, Build
Make the buy, run or host decision on evidence, and write down what it commits you to.
- Hosted, managed or local
- Run the same task hosted and local
- The model decision matrix
Method step 2, Build
Size the workload before you buy it, and find the bottleneck on paper first.
- Where the bottleneck actually is
- Estimate three small workloads
- The capacity calculator
Method step 2, Build
Build a small retrieval workflow whose answers cite sources and whose failures are visible.
- Chunking, embeddings and citations
- Build a small retrieval workflow and test it
- The evaluation set and access rules
Method step 2, Build
Containerize the application so a second person can run it without asking you anything.
- Environments, configuration and secrets
- Containerize the application and run a health check
- The container starter and config example
Method step 3, Use
Put the service somewhere real, then prove that only the right requests get through.
- Authentication, limits and least privilege
- Deploy, then test allowed and denied requests
- The deployment checklist
Method step 4, Measure
Instrument the service so you can tell an application failure from a model failure from a bill.
- Logs, metrics, traces and evaluation
- Instrument the service and build the dashboard
- Service objectives and alert thresholds
Method step 5, Share
Assemble the whole system, break it once on purpose, and hand it over.
- Production-minded without the ceremony
- Deploy, demonstrate, and recover from one failure
- The operator runbook
Requirements
- A computer you can install software on, and an account on one approved cloud or local runtime.
- Basic comfort with files and folders. Command-line work is taught in context.
- Three hours a week for four weeks.
- One document set you are allowed to index. A sample set is supplied.
About this course
The market for this subject runs from a superficial introduction to a sixty-hour infrastructure catalogue. This course takes the middle: the minimum complete system, taught visually and lab-first, ending in one reproducible deployment you can point at.
Eight modules follow a single request from the interface through the model, the data, the container, the endpoint and the dashboard, and then out to the runbook. Every module produces one working artifact, and each artifact is a component of the capstone service.
Kubernetes, multi-cloud, cluster design and GPU tuning are deliberately held for an advanced sequel. Carrying them here would trade the one thing a beginner needs, a service that actually runs, for a tour of things they will not operate this year.
The capstone is a small document assistant or task-specific service that another operator can deploy, test, monitor and stop from your documentation. It ships with an architecture diagram, a hosted against local decision, a capacity and cost estimate, an evaluation set, access test results, alert thresholds and a rollback path you have demonstrated.
Certification
Complete the course. Earn your certificate.
Complete every lesson to earn a shareable certificate of completion.
Your course certificate
- Your name
- Your full name.
- The course
- Applied AI infrastructure fundamentals
- Verification
- Verifiable online.
- Completion date
- When you finished.
Common questions
Privileged access has no course fee. Create one account to apply for Hybrid Filmmaking or join the waitlist for another course.
A course completes when your workflow runs live and you have measured it. You record a baseline in module 1, launch the workflow into a working environment, then measure the same thing again. That pairing is what makes the completion record worth sharing.
Hybrid Filmmaking accepts applications for the current intake. The other courses accept waitlist registrations for next month's intake.
Name, role, organization, and a few questions about the workflow you want to improve. It takes about a minute, and it is what lets the guided labs use your own workflow as the project.
One workflow you own, permission to change it, and the tools your team already uses. Applied AI infrastructure fundamentals also assumes a computer you can install software on, hybrid filmmaking assumes a phone or camera, and AI starter for small business runs on everyday business tools. Every course supplies sample data for the labs where your own material is restricted.
A one-page record inside the course. Before you launch, you note a baseline on time, cost, or quality. After you launch, you measure again. The sheet holds both numbers and the difference between them.
Instructor
Roan Weigert
Records all five courses
Hackathon judge and AI content creator, and host of the AI Insights San Francisco podcast. Writes this curriculum and records every core lesson.
Other courses
The other four courses
Each one takes a different kind of work through the same five steps. Pick the one closest to your job.
One permissioned campaign-to-pipeline workflow, built on your own data and measured after launch.
Combine traditional filmmaking craft with controllable AI production, then finish a hybrid film with a documented workflow.
Teaching and learning decisions first, then the policies, data rules and reviews that make them safe to adopt.
Two safe AI workflows in daily use within thirty days, with the business result written down.
Join the waitlist
See the whole AI system
Join now and we will hold your place for the October intake.