Interview Brazilian candidates in your language, at your scale.
Aya interviews every applicant in the language the role is posted in and returns a scored evaluation with the exact quote behind every score — written in the language your team reads. A team in New York can work through three hundred applicants in São Paulo without an agency, a recruiter's calendar or a single shared timezone.
Role languages available today: Português (Brasil) · Português (Portugal) · English · Español · Deutsch · Français.
Three steps, and none of them is a meeting
Post the role in your language
When you create a role you pick its language and its country. That choice is not cosmetic: it decides the language Aya opens the conversation in, the language of the application page and of every reminder the candidate receives.
Aya interviews on WhatsApp
WhatsApp is where Brazil already talks, so there is nothing for the candidate to install and no slot to book. They answer at their own pace; if they stop halfway, a reminder brings them back. A browser link is available for anyone without WhatsApp.
You read the evaluation in your language
The report is written in the language of the person who opened the role — not the language of the interview. A conversation held in Portuguese reaches your team in English, with the original quotes kept verbatim between quotation marks.
An evaluation you can defend to your hiring manager
Available in the panel, over the public API and pushed to your systems by signed webhook — the same content in all three places.
An overall score and adherence to the profile
A score from 0 to 10 plus how closely the person matches the profile you defined, as a percentage.
A score per competency, with the weight you set
You define the competencies and their weights when the role is created. The evaluation reports each one separately, so a strong candidate with one gap reads as exactly that.
The literal quote behind every score
Each competency carries the excerpt from the candidate's own answer that justifies its score, and it is flagged when we could locate that excerpt verbatim in the transcript. When we could not, it says so and asks you to confirm in a human conversation.
Points of attention and highlights
Flags never reject anyone on their own. They tell your team what to probe in the human interview.
A confidence label
Computed from objective signals in the conversation — evidence coverage, number of exchanges, richness of the answers. It is not a number the model was asked to invent about itself.
The full transcript
Everything the candidate wrote, in the original language, so nothing about the score has to be taken on trust.
Brazil is hours away from your office. The interview does not care.
Nothing is scheduled
The interview is a conversation the candidate has when they can — on the bus, after their current shift, at midnight their time. Nobody from your team joins, so the distance between São Paulo and San Francisco stops being a constraint on who you can consider.
Interviews run in parallel
Three hundred applicants do not become three hundred slots. They become three hundred conversations happening at once, and a ranked list with evidence the next morning. The ceiling is your plan's interview allowance, not a person's working day.
Nobody is dropped for being slow
A candidate who starts and stops gets a reminder in the role's language, and can continue where they left off. Half-finished interviews are the most expensive thing to lose in a funnel this wide.
What we can state, and what we will not claim
Talpy is a Brazilian company and operates under the LGPD (Lei Geral de Proteção de Dados, Brazil's general data protection law). We are not going to tell you on a marketing page that we are GDPR certified or covered by an adequacy decision, because that is a claim with legal weight and we are not making it here. If your legal team needs to assess a transfer basis, that is a conversation with us — and one we are happy to have.
What we can state, because the product does it and you can verify it in the product:
- The candidate is told they are talking to an AI in the very first message, before they answer anything.
- The candidate can ask for a human review of the evaluation. That request raises an event in the system, so it reaches your team instead of sitting in an inbox.
- The final decision is always human. Aya recommends and evidences; no candidate is approved or rejected automatically.
- Scores rest on what the person demonstrates, not on age, gender, appearance or which school they attended. There is no inference of emotion, personality or “culture fit” from tone of voice — fragile ground, and legally risky ground.
- Every action on a candidate's record is written to an audit trail: who invited them, who evaluated, who downloaded the CV or the report, and when. To be precise, because the distinction matters to a reviewer: merely opening a record in the panel is not itself an event — what gets logged is what acts on the record or takes data out of it.
- Candidates hold the full set of rights our Privacy Policy sets out — access, a copy of their data, correction, portability, deletion and withdrawal of consent — and the AI-assistant channel never exposes phone, e-mail or CV at all.
Those are the rights that matter under both regimes — which is the point of listing the rights instead of a badge. One detail worth knowing: the role's country decides which law the candidate-facing text names. A role in the EU names the GDPR to the candidate; a role in the United States names no statute at all, because there is no single federal one and citing the wrong law is worse than citing none. Our processing partners are named in the Privacy Policy, including the ones that process data outside Brazil in order to run the process you started. One more thing we would rather you hear from us: there is no automatic deletion deadline today. Candidate data stays until the company deletes the process, closes the account or the person asks for removal.
What Talpy is not
The fastest way to waste your time would be to let you assume any of these.
- We are not an employer of record. Talpy screens and evaluates; hiring, contracts, payroll and taxes are yours or your EOR's.
- We are not a staffing agency and we do not have a bench. Candidates are the ones who apply to your role, through your own link or your ATS.
- There is no certified English proficiency score. If the role needs English, make it a competency in the rubric — you get the evidence, not a language certificate.
- Plans are billed in Brazilian reais (BRL). If your finance team needs invoicing in another currency, ask us before you plan around it.
- No decision is automatic. Aya recommends and evidences; approving or rejecting is always an action taken by a person on your side.
The ones that actually get asked
Can the interview really run in English if the candidate is Brazilian?
Yes, and here is the exact behaviour, because the nuance matters. Aya opens in the language the role is posted in. From the candidate's first reply she mirrors the language they actually write in — so a candidate who answers in Portuguese continues in Portuguese. Your report still arrives in your language either way. If English is a requirement rather than a preference, put it in the rubric as a competency and the evaluation will show you how the person handled it, with quotes.
How many candidates can we screen at once?
Interviews do not queue behind a recruiter's calendar — they run in parallel, at whatever hour the candidate opens WhatsApp. The practical ceiling is the interview allowance on your plan, not a person's working day.
Do we have to move off our ATS?
No. Aya plugs in as a stage: there is a native two-way integration with Workable, signed webhooks for any other ATS, and a versioned public API. See the integrations page for what exists and what does not.
Who sees the candidate's phone number and CV?
People on your team, in the panel, and what they do there — invite, evaluate, download the CV or the report — is recorded in an audit trail. Our AI-assistant channel (MCP) is read-only and deliberately never returns phone, e-mail or CV: a score with a name is enough to decide who advances, and a phone number is not.
What does the candidate know?
That they are talking to an AI, from the first message, before they have written anything. They are told what is being evaluated, and they can ask for a human review of the evaluation. That request fires an event in the system so it does not quietly get lost in an inbox.
Tell us the role and where your team sits
Say what you are hiring for, roughly how many people, and which language the evaluation needs to arrive in. We reply with the concrete path — and if there is no path, we say that instead of booking a call.
We use this only to reply to you. Details in the Privacy Policy.