Skip to content

HR tech · Recruitment · Development

Hireloom — email-first applicant tracking system

An applicant tracking system built on the mailbox a team already uses — connecting over IMAP and SMTP so every emailed CV becomes a tracked, routed application, with AI screening you can defend and strict tenant isolation.

Project

Multi-tenant SaaS · global

Incremental delivery

Discuss a similar brief
Hireloom — email-first applicant tracking system

The challenge

Where things stood.

For most small and mid-size teams, hiring does not happen in software — it happens in a shared inbox, where CVs arrive as attachments, forwards fly around, and a spreadsheet tries to keep score. Candidates fall through the cracks, there is no single source of truth, and screening by hand is slow and inconsistent.

The tools meant to fix this make the first move the hardest: change your DNS, adopt a vendor's address, migrate your data, retrain the team. For a small company, that first move is the entire barrier to adoption.

How we built it

Delivered in deliberate phases.

  1. 01

    Email-first ingestion

    Hireloom connects to the mailbox a team already owns over IMAP and SMTP — no DNS change, no new vendor address, no migration. A background process turns new CV emails into tracked applications, and replies go out over the team's own SMTP so they stay authentically theirs.

  2. 02

    A deterministic routing ladder

    Every inbound email is resolved to a specific job — or to a review queue — by an ordered ladder of explicit rules rather than an opaque classifier, with a triage safety net at the bottom so no application is ever silently dropped.

  3. 03

    Fail-closed multi-tenancy

    Tenant isolation is enforced in the database with PostgreSQL row-level security as a backstop to application-level scoping, so a query that forgets its organisation filter returns nothing rather than another company's candidates.

  4. 04

    AI screening you can defend

    Screening returns a structured, evidence-backed result with a quote for each requirement, four autonomy modes from advisory to bounded automation, parse-confidence gating so a bad CV extraction is never auto-rejected, and a full audit record — designed around NYC Local Law 144, the EU AI Act and GDPR.

  5. 05

    An async, durable architecture

    Slow, external work — mailbox polling, parsing, model calls, sending — runs in a background worker behind a durable queue, so the interface stays fast whatever the outside world is doing, and long-running work survives restarts.

What we delivered

The work.

  • Mailbox connection over OAuth or app password, with credentials encrypted at AES-256-GCM
  • An inbound routing ladder with per-job addresses, subject codes and a triage review queue
  • Fail-closed tenant isolation using PostgreSQL row-level security
  • Requirement-based AI CV screening with per-requirement evidence, four autonomy modes and parse-confidence gating
  • Hosted careers pages and a drag-and-drop application form builder
  • Reusable templates, stage-based automations, analytics, a public API and webhooks

Outcome

What changed.

  • A team can be live in minutes with no DNS change, no new address and no data migration
  • No candidate is lost — every application is captured, routed and safety-netted by triage
  • Screening decisions are explainable, evidenced and recorded, ready for careful buyers and regulators
  • The mailbox stays authentically the team's, so the candidate experience does not change
  • The platform scales from a first hire to high-volume campaigns on one system

Meet hiring where it already happens — the inbox — and build the system of record around it, rather than asking a team to move somewhere else just to get any value.