
HairaOur own AI recruitment platform. It ranks resumes and runs AI phone screens and AI video interviews, and went from first commit to a live, paid product in about 6.5 months.
The challenge
Hiring one role can bring in hundreds of applications. The first steps repeat every time and eat most of a recruiter's week: opening every resume, phoning candidates to check notice period, salary, city and interest (with plenty of unanswered calls and phone tag), and booking 15 to 30 minute first interviews around two calendars.
The result is slow hiring (good candidates take another offer while they wait), inconsistent judgement (different recruiters ask different questions, and tired people judge differently late in the day) and wasted cost, because agencies are paid per placement. Existing applicant tracking systems store candidates and move cards across a board, but a person still does every screening step by hand.
Our solution
We designed and built Haira, a multi-tenant applicant tracking system with AI phone screening and AI video interviews built in, for recruitment agencies and in-house talent acquisition teams, India first and then international. AI does the first three rounds (resume review, phone screen and first interview) for every applicant, the same way each time. The recruiter's job shifts from doing the screening to reviewing its results.
Under the hood: a FastAPI and Python backend with an ARQ worker, a Next.js 16 and React 19 web app, three LiveKit voice agents (video interview, phone screen and a public demo), PostgreSQL, Redis, MinIO, Qdrant and Meilisearch, with each AI task matched to a model chosen for cost, speed or accuracy. The model suggests and the code decides: scores are clamped, capped and backed by quotes from the resume, so every number can be explained.
- Candidates apply from a public job page, the company careers widget, LinkedIn, Google for Jobs or job boards
- Each resume is scored out of 100 in seconds, with rules tuned per industry
- Two company-set cut-offs reject, hold or advance each candidate automatically
- An AI voice agent phones the candidate in office hours for a five-minute screen
- An AI gatekeeper recommends advance or reject, with a reason
- A personalised AI video interview in the browser, any time within 48 hours
- A scored report on technical skill, job fit, communication and confidence
- The recruiter takes over in a Kanban pipeline through to offer and hire
Key features
- Resume ranking in 8 sections with evidence quotes, skill caps and code-computed employment gaps
- AI phone screening over LiveKit SIP, on a shared Plivo line or the company’s own number
- Redial after a dropped call that continues where it stopped, up to 3 attempts per cycle
- AI video interviews with 5 to 12 personal questions, fullscreen and camera integrity checks
- Explainable four-dimension assessment with "Not assessed" instead of a zero
- Protection against hidden-text prompt injection in resumes, with an integrity verdict
- Two-layer access control and strict tenant isolation, including the AI assistant
- Haira Assistant: plain-English questions over the company’s own candidates
- Job publishing to Google for Jobs, LinkedIn, social channels, 7 aggregators and an embeddable careers widget
- Credits that are held, then settled, with Razorpay billing in INR (with GST invoices) or USD
The results
Haira is a live, paid product covering the full flow from job post to hire, with three AI rounds that run without a recruiter. Phone screening replies in about 2.2 to 2.5 seconds, video interviews produce an explainable score, and the commercial layer (plans, credits, holds, two currencies and tax invoices) is complete. A full security review before launch was completed, with every code-level finding fixed within a day.
Targets such as "10× faster recruiting" and "60% shorter time-to-hire" still need customer data, so we do not claim them yet. Read the full story, including the architecture and the hard problems we solved, in our Haira case study, or try the product at haira.ai.
- commits from 4 May to 18 November 2026
- ~2,860
- commits from 4 May to 18 November 2026
- merged pull requests and 243 releases
- 1,021
- merged pull requests and 243 releases
- Python test functions, plus 203 frontend test files
- ~4,080
- Python test functions, plus 203 frontend test files
- written feature specs
- 155
- written feature specs
Screenshots




Client
The Special Character (own product)
Industry
HR tech: AI recruitment software
Timeline
About 6.5 months to launch (May to November 2026)
Team
27 contributors


