Interview case study generator with dummy datasets
Build an internal tool that turns a job kit into interview take-home case studies and archives every one that's been finalized. Who it's for and why. Hiring managers upload a job kit PDF, describe what they want the case study to test for, and get a few AI-drafted take-home case studies to choose from — each scoped to a realistic few-hour build and ending with a short recorded walkthrough of the deliverable and the candidate's reasoning. Choosing one produces a candidate-facing brief with a shareable link (plus a dummy dataset file when the case study needs one), and every finalized study is archived for reuse. Trigger. No schedule or webhook — everything starts from a person opening an internal web app. Build a studio app on a private route (space members only) as the entry point. Main flow. A "new case study" wizard where the user uploads the job kit PDF and describes what they're testing for. Extract plain text from the PDF, then kick off a generation job that drafts a few distinct case-study options. Let the user pick an option, or leave feedback for a single regeneration pass. On finalize, capture a role title and team, generate the dummy dataset if the chosen option calls for one, save the record to permanent storage, and mint a candidate share link. A repository view listing every finalized case study. A candidate-facing page, opened via each study's own shareable link, that shows just that one brief (printable to PDF) — reachable by people outside your org with no account, and scoped so a link only ever exposes its own record. Async generation (important). The AI drafting calls are too slow to run inside a synchronous browser request (gateways typically time out long-running requests). Use an enqueue → background worker → poll pattern: an enqueue step validates the request, writes a pending job, and returns a job id instantly; a linked worker step runs after the response is sent, calls the model, and writes the result; the app polls a lightweight job-status endpoint until the job is done or errored. Apply this to both the option generation and the finalize/dataset generation. Storage. Use persistent storage for two things: a permanent repository of finalized case studies (each owning its own record, metadata, and optional dataset file), and transient async-job records that can be pruned after they age out. AI. Use an LLM (I used Anthropic's Claude — or your preferred model provider) via a connector for both drafting the options and generating the dummy dataset. This is the only external service; everything else is internal. Tools used: Anthropic Claude API, 3B
What this prompt builds
An internal tool that converts job kits into AI-drafted interview case studies with accompanying dummy datasets. Hiring managers upload a job description, specify what to test for, and receive multiple case study options complete with candidate-facing briefs and shareable links. The workflow archives every finalized case study for reuse across the team.
The problem
Hiring managers in commercial operations and RevOps teams spend multiple cycles designing realistic interview case studies and generating the dummy data candidates need to complete them. Creating case studies that accurately reflect real projects while remaining scoped to a few-hour exercise is time-intensive, and building accompanying datasets from scratch compounds the effort. This workflow turns a job kit PDF into three AI-drafted case study options, each scoped to a realistic take-home exercise and paired with a generated dummy dataset when needed. Managers describe what they want to test for, review the options, and finalize one to produce a candidate-facing brief with a shareable link—returning hours to the hiring team while arguably producing better, more consistent case studies.
Solution and impact
The workflow gives RevOps and hiring teams hours back by automating case study creation and dataset generation. Every finalized case study is archived in a searchable repository for reuse across roles and teams, reducing redundant work. Candidates receive polished, consistent briefs via shareable links, and hiring managers can iterate on AI-drafted options with a single regeneration pass before sending.