Capstone: design a grounded career helper | Introduction to Generative AI
Understand how generative AI works, write source-bounded prompts, validate outputs, evaluate grounded answers and design a responsible career-learning assistant.
Prerequisites: No prior AI experience required. Basic familiarity with reading short code examples is helpful; no paid API account is needed.
Capstone: design a grounded career helper
Your capstone is a design and evaluation package for a small career-learning assistant. Choose a narrow question-answering task, such as explaining the learning requirements for a course. Supply an approved set of short policy passages and define how the assistant should behave when information is missing. Build a prompt and a response contract that includes an answer, supporting passage IDs and an explicit uncertainty field.
Create a test set before presenting a successful demo. Include at least six examples: two normal questions, a question with no supporting passage, conflicting passages, an instruction embedded inside source data and an attempt to request another candidate personal information. For each case, specify the expected behavior and the evidence a reviewer should inspect. It is acceptable for the prototype to abstain; an honest limitation is preferable to an unsupported claim.
Generative systems can produce plausible but false material. The NIST generative-AI risk profile discusses this risk as confabulation and emphasizes evaluating systems in their intended context. For this project, translate that concern into concrete checks: require source support, protect personal data and make failures visible to the user. Read the primary reference linked below for a broader risk-management perspective.
Finish with a short README containing the task, data sources, prompt version, response schema, test cases and known limitations. Review your own work against the assignment rubric. Completing these reading lessons and passing each chapter assessment makes you eligible for a JobsSearchIndia course-completion certificate. The project is a self-review exercise; the platform does not execute, independently grade or accredit your code.
Worked example
{
"answer": "Complete every lesson and pass every chapter assessment.",
"source_ids": ["certificate-policy"],
"uncertainty": "This answer only describes the supplied course policy."
}
Try it yourself
Prepare the capstone README and evaluation table. Submit your design in the Projects section for saved self-review, then complete the chapter quiz. Do not include credentials or real candidate personal information.
Check your understanding: explain why the example works, state one limitation, and record how you would test it.
NIST AI 600-1: Generative Artificial Intelligence Profile
1. Understand generative AI
2. Give clear instructions
3. Ground and evaluate answers
4. Build a responsible learning prototype