AIDA Portal
About Us

Watch · About a minute

AIDA Portals, in about a minute each

For job seekers

For recruiters

Captioned, no sound needed.

AIDA Labs · Product

AIDA Portal

Artificial Intelligence Data Analysis, delivered through a secure portal. An AIDA Portal takes what a client gives it, runs it through AIDA's analysis, has a person review the results, and publishes them to a portal the client signs into. Two AIDA Portals are running today, on one shared platform.

Two portals, one platform

Job Seeker Portal

aida-job-search.ca ↗
Live, in pilot

For job seekers. Instead of thousands of search results, a short, honest list of real, currently open jobs, each scored against the client's background and reviewed by a person before the client sees it.

  • A keyword matrix built from the client's background, in seven weighted layers: title, skill, tool, certification, industry, seniority and location
  • Live postings collected every day from company career pages, Canadian Job Data and employer hiring-system feeds, for Canada and (by choice) the US
  • Every posting scored against the matrix by one database function, with whole-word matching
  • The Top 100 reviewed by a person (Data Queen): Best Match, Worth a Look or Removed, each with a written reason
  • Ghost-job and employer checks: stale reposts and employers with a poor track record are flagged
  • Scheduled rescans, daily to monthly; after the first review, only new postings need a fresh look
  • Deliverables: Jobs Report, Career Bio, Search Strategy, resumes (Word, PDF and ATS text), cover letters, application kits for one specific job, interview playbooks
  • In the portal: picks with their reasons, Request a Scan, Not Interested, resumes and kits

Recruiter Portal

recruiter.aida-job-search.ca ↗
Being configured

For recruiters ("Know Your Candidate"). The job-seeker side turned around: one job, many candidates, every candidate scored against the job and explained.

  • One job description and a stack of candidate resumes
  • Every resume normalized into the same format first, so candidates are compared like for like
  • The same scoring engine as the job-seeker side; the keyword matrix drafted for the job is approved by a person before it scores real candidates
  • A human-reviewed deep pass on the strongest candidates
  • Every candidate sorted Best Match, Worth a Look or Weak Fit, with the reason why; explanations never invent missing detail
  • Candidates never get logins and can be reused across postings
  • Deliverables: candidate profiles and a candidate analysis report
  • Credits are used per deep-reviewed candidate

How an AIDA Portal works

Both portals run the same six steps. Only the intake and the deliverables change.

IntakeThe client provides the input: a resume and background, or a job description and a stack of resumes.
StandardizeEverything is put into one clean, consistent form: postings deduplicated and checked, resumes normalized.
ScoreOne database function scores every item against the keyword matrix. Plain, versioned SQL, the same for every client.
ReviewA person reviews the ranked results (Data Queen) and writes the reason for every decision.
DeliverReports and documents are built from the reviewed results, to fixed formats, with no invented detail.
PublishResults sync from the data platform to the client's portal on a separate server; the client signs in to see them.

Data collectors: dozens of sources, one clean format

A portal is only as good as the data behind it. AIDA designs and builds data collectors that pull public information from dozens of sites, each of which builds and presents it in its own way, and turn it into one clean, consistent record. For the Job Seeker Portal, three collectors run every day:

Company career pagesEmployers' own career sites in Toronto and Vancouver, scanned every night. Each site is built and laid out differently, so each gets its own extraction rules; long descriptions are trimmed to a standard length.
Canadian Job DataA federal job board, scanned every four hours with a real browser engine. Its "load more" button pads small searches with unrelated jobs, so the collector keeps only the results the site itself counted.
Employer hiring-system feeds51 employers on five hiring platforms (Greenhouse, Lever, Ashby, SmartRecruiters, Workable), read through their public feeds. Adding an employer is one registry entry, and every run records ok, empty, not found or error for each one.
Keeping it currentCheckers revisit postings and mark closed ones, so they drop out of scoring. A posting found by two collectors is stored once, keyed by its address.

What every collector does

  • Public information only, collected on a schedule spread through the day
  • Every posting standardized to one record: title, company, location, description, source, search term, country
  • Each posting stored once, keyed by its address
  • Each new posting scored the moment it arrives, by the same function as everything else
  • Every run logged, so a source that changes or breaks shows up right away
  • Each collector is its own workflow, so one site changing its layout can't break the others

The same approach works beyond jobs: any public information that is published across many sites in different ways.

Live job-market data

The database behind the Job Seeker Portal is open to explore, read-only. See what the collectors gather every day, where it comes from and how every posting is checked.

Jobs Data dashboardPostings added per day, sources, regions, industries, top employers, quality and link checks, and scoring totals, live from the database. Every number, bar and row opens the postings behind it, with a link to each original advert.Open the live dashboard ↗Read-only, and nobody is namedThe dashboard can run two read-only reports and nothing else. It shows public job adverts and anonymous totals, never a job seeker, client or recruiter.Details ›

Built with guardrails

AIDA uses AI sparingly: to help build the portals and prepare documents, never as an unsupervised actor on client data.

Scores come from plain SQLMatching and scoring are deterministic database functions, not a model's judgment at run time.
A person reviews every resultNothing ranked reaches a client until a person has reviewed it and written why.
No agent writes to the databaseAI prepares changes as files; a person reviews and applies every one. Loaders refuse the wrong kind of file before writing anything.
No invented detailResumes, cover letters and fit explanations use only facts the client provided.
Staff and client sides kept apartThe staff dashboard runs on the data platform behind its own access; clients only ever reach their own portal.
Cheap to runBuilt on free and open-source software; the whole platform costs about $15 a month to operate.

All ten guardrails AIDA runs on: About Us ›

Where it runs

Data platform

  • PostgreSQL database: postings, profiles, scores, reviews
  • Staff dashboard and the scoring function
  • Daily data collectors
  • Everything in containers

Client portals

  • A separate server in Canada hosts the portals
  • Results synced from the data platform; client requests synced back through a checked job
  • No client ever reaches the data platform

Network

  • DNS and HTTPS for every portal
  • A secure, outbound-only tunnel to the data platform, with no inbound ports open