Job search command center with a self-filling pipeline board

I want a job hunt command center I open every day, an app that replaces the application tracker spreadsheet I keep by hand. The whole point is that the board fills itself from live postings instead of me copying roles into rows, and that every saved card tells me what the role is actually worth.

One side of the app is live search over JobsPipe. Give me a filter panel for job title, location, remote status, a salary floor and required skills, and run Search Job Postings when I submit those filters. Show the results as a compact list with the title, company, location, remote flag, the advertised pay when the posting states it, the posting date and the source. Each result gets a Save button that puts it onto my board.

The other side is the pipeline board, with five columns: Saved, Applied, Interviewing, Offer and Closed. I drag cards between columns and the position sticks. Keep the board, the cards, my notes on each card and everything the app has enriched in the General Input Database, so the board is exactly where I left it when I come back. Boards are per user, and one person can keep several named boards, because a career coach needs to run a separate board for each client and switch between them.

Each saved card should show three pieces of context I would otherwise research by hand. First, how the pay compares to the market: use Get Salary Benchmark and Get Occupation Compensation Percentiles for the posting's occupation, and show where the advertised range sits against the p10 to p90 spread. Second, the employer's technology stack from Scan Domain Technology Stack on the company domain, so I can prepare for the interview. Third, the skills the market expects for that occupation from Get Top Skills For Occupation, compared against the skills in my stored profile so the card shows my own gaps.

Every card needs a Prep this application button that kicks off a background agent. The agent reads the full posting text, compares it against the profile and resume text I have saved in the app, then writes two documents in Google Docs: a tailored cover letter, and an interview prep sheet covering likely questions, the employer's stack and my skill gaps. Use Create Document to make each one and Batch Update Document to fill in the content. It should also stage an outreach email to the recruiter or hiring contact using Gmail Create a Draft, left unsent for me to review. When it finishes, write the two document links and the draft link back onto the card so the board shows at a glance that prep is done.

Give me a profile area where I store my profile summary, my resume text and my target occupation, since both the prep agent and the skills gap read from it. For a coach running client boards, this profile belongs to the board, so each client has their own resume and target role.

A few things to build around. JobsPipe postings are already normalized and deduplicated across 30 plus ATS and job board sources, so the same role advertised in three places arrives once and the app needs no dedup of its own beyond refusing to save the same posting twice. Only about a quarter of postings state pay at all, so treat a null salary as the normal case and fall back to the occupation benchmark, labelling it clearly as a market estimate rather than the employer's advertised figure. Job records carry a full description field so wide result pages get large: page the results using the returned cursor behind a Load more control rather than pulling everything at once. Every JobsPipe request costs one credit, including a request that returns zero results, so run searches on explicit submit rather than on every keystroke, and cache the salary, stack and skills enrichment onto the card in the database instead of refetching it each time the board loads. Upstream stack scans are cached for about two weeks, so a per card refresh button is enough. When an occupation has too few postings the insights come back suppressed with null percentiles, so show a plain not enough market data state instead of an empty chart. Always use the live search operation and never the sandbox one, since sandbox routes return canned sample data.