Airbase — Sage, the AI Recruiter
An AI-powered ATS where Sage, the AI recruiter, runs your hiring pipeline end to end.
The role
Founding Engineer — architecture & full build
Highlights
- Built Sage, an AI recruiter that runs the hiring pipeline end to end, from a configured role to a ranked shortlist, so recruiters only meet the best.
- Architected the full platform — ATS API, recruiter app, interview client, recording, async workers, a real-time AI interview agent, and a multi-channel bot — covering the whole hiring lifecycle.
- Built a real-time voice agent that autonomously conducts video and phone interviews, replacing first-round human screening.
- Built AI sourcing that finds candidates, writes job descriptions in your house style, and screens every application against your rubric.
- Built a stage-based automation engine: recruiters set rules in plain English ("advance anyone scoring 90+, email the rest"), and AI evaluates the condition and runs the actions.
- Built an in-app AI assistant that runs real tasks on command — operating the app through its own API as the logged-in user. Open-sourced as agentside.
- Cut recording and document-processing costs by ~90–99% with custom infrastructure in place of costly managed services.
- Model-agnostic LLM orchestration with automatic provider fallback, plus GDPR data-privacy workflows built in.
// inside_the_product
What Airbase does — and what I built
AI Recruiter
Sage runs your hiring pipeline
Hand Sage a role and it takes it from there, moving every applicant through screening and interviews until a ranked shortlist is ready. Recruiters stop chasing the pipeline and spend their time on the people worth meeting.
Sage ⟩ Walk me through a system you scaled.
“Sure — we moved to an event-driven…”
AI Interviews
Structured first-round interviews, on autopilot
A real-time voice agent interviews every candidate over web or phone — asking your questions, following up naturally, and scoring as it goes. I built the sub-second voice pipeline behind it.
Resume Screening
A ranked shortlist before you open a resume
Every application is read against your rubric and scored, so the strongest matches surface on their own — with structured, evidence-backed decisions, not vibes.
AI Sourcing
It sources, writes, and screens for you
Sage finds candidates, drafts job descriptions in your house style, and reads every application against your rubric, so the pipeline fills itself.
Talent Matching
Rediscover talent you already have
Search everyone who has ever applied and rank them by fit for a new role. Your next hire may already be a yes.
Omnichannel
Meet candidates where they are
Interviews by voice, video, text, or email, in 5+ languages, so you reach your whole audience and far fewer people drop off.
CRM
Companies, contacts, and your team
A built-in CRM for client and candidate relationships, with roles and permissions for the whole team, so nothing about a relationship gets lost between hires.
Workflow Automation
Automations you describe in plain English
Set rules on any stage the way you'd say them — "advance anyone scoring 90+ to Offer and email them; send everyone else a polite no." The platform turns that into an AI-evaluated condition and runs the actions on its own: move the candidate, send an email or SMS, and fire steps before and after each AI interview.
// architecture
The system at a high level
A high-level view — a TypeScript/Bun platform of focused services with AI agents embedded across the hiring lifecycle, model-agnostic and API-driven.
- 01
Clients
ReactWebSocketsService WorkerThe recruiter and applicant apps, plus a real-time interview client with client-side recording that survives flaky networks.
- 02
Core ATS API
BunPostgreSQLRESTThe system of record for the whole hiring pipeline, with team permissions and GDPR data-privacy workflows built in.
- 03
AI layer
agentsideMulti-providerTool-useSage, scoring that grades against rubrics with structured output, and an in-app agent (agentside) that operates the ATS through its own API.
- 04
Interviews & recording
STT / TTS / VADTwilioS3A sub-second voice agent that bridges browser and phone callers into one interviewer, paired with a cost-effective custom recording service.
- 05
Async & integrations
BullMQRedisSlack / TeamsQueue workers for the heavy AI jobs and a multi-channel bot for Slack and Teams.
What it is
Airbase is an AI-powered applicant tracking system, and Sage is the AI recruiter at its center. You set up a role and Sage runs the pipeline from there. It finds and screens candidates against your criteria, interviews them by voice, video, text, or email, goes back through everyone who has applied before, and keeps things moving until you have a ranked shortlist to review. Recruiters can also ask Sage to do things in the app, and it does them.
Underneath is a full ATS with jobs, stages, scorecards, a CRM, and automations, so the team runs its whole hiring process in one place. Airbase started life as two products, Cuee and Airbase, which are now one.
I'm the founding engineer. I designed the architecture and built the platform largely solo across TypeScript, Bun, and Python — a set of focused services with AI agents embedded throughout.
The hard parts
- An AI recruiter that runs the pipeline. Sage takes a role from setup to shortlist and does real work along the way, not just chat.
- A voice agent that conducts interviews. Real-time, over web and phone — it listens, asks, follows up, and scores, on its own.
- Automations in plain English. Recruiters write stage rules the way they'd say them, and the platform evaluates the condition with AI and runs the actions.
- An assistant that acts, not chats. It runs real tasks by driving the app through its own API, as the logged-in user, inside their permissions.
- Decisions you can audit. Scoring grades against rubrics and returns structured, evidence-backed output — not vibes.
- Production quality at low cost. Custom infrastructure cut recording and document costs by ~90–99%, and no single model provider can take it down.
Want something like this built?