HireHaath: a hiring app for workers who would rather speak than type
A staffing company placing drivers, warehouse staff, electricians and delivery partners was losing candidates at every step of a form-heavy process. We built HireHaath: a voice-first worker app in Hindi, AI profile building from a short conversation, employer job posting and matching, and interview scheduling over WhatsApp — 25 features designed for workers, not HR departments.
Candidates who vanished halfway through a form
HireHaath's founder showed us their funnel. Thousands of workers downloaded the app every month; a small fraction finished their profile. The profile form had been designed like a white-collar job portal: education history, skills checkboxes, a résumé upload. For a delivery partner or a warehouse loader, most of it was irrelevant — and typing it on a small phone in English was a barrier in itself.
Employers had the opposite complaint: the few profiles that were complete often lacked the details that mattered to them, like which licence a driver held or whether an electrician had worked on industrial wiring.
“We were asking people who build our cities to write a CV. They just want to tell someone what they can do.”
Listening to workers
We interviewed workers at labour chowks, warehouses and a driver training centre. Almost everyone was comfortable talking about their work — for minutes, in detail — and uncomfortable typing about it. Many used WhatsApp voice notes daily. Trust was a concern too: workers had been charged fees by fake agents and wanted to know that a job was real before travelling for an interview.
Employers, in turn, wanted a short list of relevant, nearby candidates with the right documents — not hundreds of applications.
- Speaking is easy; typing in English is not
- Distance to the job matters more than almost anything
- Fear of fake jobs and agents who charge fees
- Interviews missed because details were unclear
A profile built from a conversation
The worker app asks a few simple spoken questions in Hindi — what work do you do, for how long, where do you live, what licences or certificates do you have. HireHaath transcribes the answers and an AI model turns them into a structured profile. The worker sees the profile read back to them, with big buttons to confirm or correct each part.
A driver can complete a useful profile in two minutes without typing a word. Documents like licences are captured by camera and checked for readability.
The team
Voice and matching were the core risks, so AI engineering was staffed from the first sprint.
Worker and employer research, launch experiment.
Voice-first worker experience and employer portal.
Worker app, voice capture and document camera.
Employer portal and recruiter console.
Jobs, applications, scheduling and WhatsApp channel.
Hindi speech-to-text, profile extraction and matching.
8 people in total, working as one team.
Decisions we made
These were made with the founder and the recruiter team, who knew the workers best.
Typed forms or voice-first profiles?
- Simplified typed forms
- Voice-first profiles with confirm-and-correct
Our call: Voice-first profiles with confirm-and-correct. Workers could describe their work in detail when speaking. Voice removed the biggest barrier to completing a profile.
How to match workers to jobs?
- Keyword search by employers
- Matching on skills, distance, shift and documents
Our call: Matching on skills, distance, shift and documents. Distance and shift fit predicted whether a worker would actually show up. Matching on them produced short, relevant lists.
How to fight fake jobs?
- Open job posting
- Verified employers only, no fees charged to workers
Our call: Verified employers only, no fees charged to workers. Workers had been exploited by fake agents. Verifying every employer and never charging workers built the trust the platform needed.
App notifications or WhatsApp for interviews?
- Push notifications
- WhatsApp messages with maps and reminders
Our call: WhatsApp messages with maps and reminders. Workers checked WhatsApp constantly and often cleared app notifications. Interview details on WhatsApp reduced no-shows.
The 25 features
Everything that shipped for workers, employers and HireHaath's recruiters.
- 01Voice profile builder
Answer spoken questions in Hindi; get a structured profile.
- 02Read-back and correct
Profile read aloud with big confirm or change buttons.
- 03Document camera
Licences and certificates captured and checked.
- 04Nearby jobs
Jobs sorted by distance, shift and pay.
- 05One-tap apply
Apply without filling another form.
- 06Interview details on WhatsApp
Time, address map and recruiter contact.
- 07Job posting
Role, location, shift, pay and requirements.
- 08Ranked matches
Candidates ranked by fit and distance.
- 09Voice introductions
Listen to a candidate describe their experience.
- 10Interview scheduling
Slots booked and confirmed automatically.
- 11Hiring pipeline
Track shortlisted, interviewed, hired and joined.
- 12Hindi speech-to-text
Accurate transcription tuned on real worker speech.
- 13Profile extraction
Skills, experience and licences structured from speech.
- 14Matching engine
Skills, distance, shift and document fit.
- 15Duplicate profile detection
One worker, one profile.
- 16Employer verification
Business documents checked before posting.
- 17No-fee guarantee
Workers never pay; reported violations investigated.
- 18Recruiter console
Recruiters assist workers and employers.
- 19Placement tracking
Joining and retention after placement.
- 20Role-based access
Workers, employers, recruiters and admins.
- 21Consent and privacy
Workers control what employers can see.
- 22Analytics
Profiles, matches, interviews and placements by city.
- 23Fraud reports
Workers can report suspicious jobs in one tap.
- 24Multilingual expansion
Language packs for new regions.
- 25Audit trail
Every verification and moderation action recorded.
Launch
We launched the voice profile to a portion of new sign-ups first and compared completion with the old form. Once the difference was clear, voice became the default for everyone. Interview no-shows dropped after we moved interview details and reminders to WhatsApp, with a map link and the recruiter's name.
- Weeks 1–2Discovery
Worker interviews at chowks, warehouses and training centres.
- Weeks 3–5Voice prototype
Voice profile tested with workers in Hindi.
- Weeks 6–13Build
Worker app, employer portal, matching and WhatsApp.
- Weeks 14–16Launch experiment
Voice profile against the old form for new sign-ups.
- Weeks 17–18Full launch
Voice becomes the default for all workers.
What we learned
Design for how people naturally communicate. Voice unlocked a user group that forms had excluded.
Trust is part of the product. Verified employers and clear job details mattered as much as matching quality.
- React Native
- Next.js
- NestJS
- PostgreSQL
- Hindi speech-to-text
- LLM profile extraction
- Matching engine
- WhatsApp Business API
