Machine Learning Engineer (Remote)

About Kinterview: Kinterview (kinterview.cc) is a real-time AI interview copilot. It listens to interviewers' questions and streams tailored answers to a private window, in 23 languages, on Zoom, Google Meet, and Microsoft Teams. We're a remote-first team on a mission to end interview anxiety for job seekers everywhere.

Location: Remote, work from anywhere (some overlap with UTC+8 mornings preferred).

What we offer: Competitive salary; fully remote and async-first; annual team offsite; learning & equipment budget; early-stage equity.

How to apply: Email your resume and a short note to [email protected] with the role title in the subject line.

01What you'll do

• Improve answer quality: prompt pipelines, retrieval over resumes and job descriptions, evaluation harnesses.
• Reduce latency of the real-time transcription-to-answer loop.
• Build evals that catch regressions before users do.

02What you bring

• 3+ years with LLM systems in production: prompting, RAG, evals, inference optimization.
• Strong Python; comfortable with streaming architectures.
• You measure quality, not vibes.

03Nice to have

• Speech recognition / diarization experience.
• Multilingual NLP experience.

04About the product you'll be working on

Kinterview is a real-time AI interview copilot. When a candidate is in a live interview on Zoom, Google Meet, or Microsoft Teams, Kinterview listens to the interviewer's questions and streams tailored answers to a private window, in 23 languages.

Interviews are high-stakes moments, and the product exists to support candidates in that moment: the transcription has to keep up, the answers have to fit the question being asked and the candidate's own background, and the interface has to stay calm and out of the way while the candidate is under pressure. Every part of the product serves that moment.

05How the team works

We are a remote-first company and a small team (1-10 people), hiring remotely worldwide. We work async-first and in writing across time zones, with some overlap with UTC+8 mornings preferred so we can meet live when it matters. There is no handoff theater: the people who design and build a feature are the people who hear from users about it, and everyone owns their work end to end, from idea to shipped, measured, and iterated. If you like small teams where your work reaches users directly, you will fit in here.

06What we offer

Competitive salary, fully remote and async-first work, an annual team offsite, a learning and equipment budget, and early-stage equity.

07How to apply

Email your resume and a short note to [email protected] with the role title in the subject line. Tell us briefly why this role and what you have built or done that is most relevant.

More on this role: your day-to-day work centers on the loop from live transcription to a tailored answer in the candidate's private window. That means working on the prompt pipelines that shape what the model produces, on retrieval over resumes and job descriptions so the answer fits the person and the role in question, and on the evaluation harnesses that tell us whether an answer is actually good.

You will spend time reducing the latency of that transcription-to-answer loop, because in a live interview a slow answer is a missed answer, and on building evals that catch regressions before users do, so a change that looks fine on one example cannot quietly make answers worse across the range of interviews the product supports. In production, that work draws on prompting, retrieval-augmented generation, evaluation, and inference optimization, written mainly in Python and running over streaming architectures.

The standard here is to measure quality rather than go on vibes: if a change matters, it should show up in an evaluation, and if it does not, we should know that too. Your background in speech recognition and diarization will matter where the transcription itself has to keep speakers and turns straight, and multilingual experience will matter because the product works in 23 languages, the same quality bar applies however the interview is conducted.