Save on courses or Exam now
About Instructors
-
tomyRoger- Instructor Rating - Reviews - Students - CoursesNo biography available, please complete
📘 About This Course
Design AI interfaces people can actually use - and prove it.
Most "add an AI assistant" projects fail in the interface, not in the model. Users cannot tell what the system is able to do, they cannot tell when it is wrong, and nobody can say whether the experience actually improved. This course closes that gap. Across ten modules and 58 lessons you design, build and validate a complete human-AI interface, working in a real Linux lab against a real multimodal LLM REST API.
You start by setting up a safe Linux workspace, inspecting the API's verified capabilities and limits, and opening a research repository that links user evidence, model behaviour and design decisions. From there you work through the UX foundations that are specific to AI: interaction patterns and mental models, affordances and expectation setting, levels of autonomy, progressive disclosure of uncertainty, and a repeatable heuristic critique.
Conversational design comes next - turn architecture, prompts treated as interface, onboarding and starter prompts, clarification and recovery flows, context and memory controls - followed by multimodal interaction: choosing the right modality, image and document input with consent, cross-modal referencing, latency and streaming states, and accessible alternatives when a modality is unavailable.
Then you build. In the Linux lab you turn a user flow into a running web prototype, connect it to the multimodal REST API without exposing credentials, implement conversation state and reset controls, add image input with preview, and instrument privacy-conscious UX events.
The second half is validation. You write a research plan that tests both usability and the usefulness of AI output, run task-based sessions without coaching participants, score outputs with user-centred rubrics, code think-aloud notes, run controlled comparisons, and turn findings into prioritised design decisions. You define UX and AI quality metrics together, design error taxonomies and recovery, work on trust and appropriate reliance, and assemble a release-gate scorecard. Accessibility, inclusive language, bias across user segments, privacy and consent, unsafe-output friction and dark-pattern avoidance each get their own lesson, followed by design systems, specification and API-contract handoff, acceptance criteria and design QA.
The capstone ties it together: you brief, build, test, fix and defend an end-to-end multimodal human-AI interface, and close with a launch recommendation supported by reproducible evidence instead of design opinion.
Everything is demonstrated, not described. Every lesson shows the real terminal, the real browser, the real API responses and the real prototype, including a controlled failure that is diagnosed, corrected and retested. Every lesson ships with its PDF support deck and the code that was actually executed on screen.
What You'll Learn
Design human-AI interfaces whose capability cues match what the model can really do.
Build a working multimodal prototype on Linux against a real LLM REST API, without exposing credentials.
Run task-based usability tests and validate model output with user-centred rubrics.
Communicate uncertainty, errors and recovery so users rely on the AI appropriately.
Keep dynamic AI states accessible and inclusive, and avoid manipulative patterns.
🛠️ Requirements
- Comfort with a Linux terminal (SSH, Bash, editing files) and basic Python or JavaScript.
- A Linux machine or VM you are allowed to use, plus access to a multimodal LLM REST API. No prior UX research experience needed.
👤 Who This Course is For
- UX and product designers, and the developers who build AI features into a real product.
- Researchers and product managers who must validate AI behaviour with users and decide whether to ship.
53618 Students applied
100 Comments
-
-
John DoeMorbi nec nisi ante. Quisque lacus ligula, iaculis in elit et, interdum semper quam. Fusce in interdum tortor. Ut sollicitudin lectus dolor eget imperdiet libero pulvinar sit amet
-
John Doe
Morbi nec nisi ante. Quisque lacus ligula, iaculis in elit et, interdum semper quam. Fusce in interdum tortor. Ut sollicitudin lectus dolor eget imperdiet libero pulvinar sit amet
-
-
John Doe
Morbi nec nisi ante. Quisque lacus ligula, iaculis in elit et, interdum semper quam. Fusce in interdum tortor. Ut sollicitudin lectus dolor eget imperdiet libero pulvinar sit amet
-
John Doe
Morbi nec nisi ante. Quisque lacus ligula, iaculis in elit et, interdum semper quam. Fusce in interdum tortor. Ut sollicitudin lectus dolor eget imperdiet libero pulvinar sit amet
-
John Doe
Morbi nec nisi ante. Quisque lacus ligula, iaculis in elit et, interdum semper quam. Fusce in interdum tortor. Ut sollicitudin lectus dolor eget imperdiet libero pulvinar sit amet
Morbi nec nisi ante. Quisque lacus ligula, iaculis in elit et, interdum semper quam. Fusce in interdum tortor. Ut sollicitudin lectus dolor eget imperdiet libero pulvinar sit amet