ML Engineer, Voice & Document AI
Improve the accuracy and speed of Atlas and Echo, our document and voice platforms, on Japanese documents and speech.
- Team
- Applied ML
- Location
- Tokyo · Hybrid
- Employment
- Full-time
- Experience
- 3+ years
- Languages
- Business-level English
What you’ll do
- Build and evaluate extraction for Japanese documents: tables, vertical text, stamps and handwriting.
- Tune streaming speech recognition and synthesis for low-latency Japanese voice agents.
- Create datasets, labelling workflows and evaluation sets with client teams.
- Benchmark and select models, including open-weight ones we host ourselves.
What we’re looking for
- 3+ years of machine learning in production.
- Strong Python and PyTorch.
- Rigorous evaluation habits and comfort with messy real-world data.
- Experience building evaluation pipelines for non-English or multilingual NLP or speech systems.
Nice to have
- Business-level Japanese.
- OCR or document layout models.
- Speech recognition or synthesis.
- GPU inference optimization.
Hiring process
Four rounds, and a clear answer at every step.
Interviews are held online, in English or Japanese.
- Start
Apply
Email your resume and a cover letter to career@kontextly.ai. We read every application and reply within 5 business days.
- Round 130 minutes
Initial screen
A short AI interview covering your background, experience and motivation, so we can quickly find a good match on both sides.
- Round 290 minutes
Coding round
A practical exercise close to the work we do, in your own setup. No puzzles or trick questions. Use the tools you normally use, including AI assistants, and talk us through your choices.
- Round 360–90 minutes
Technical interview
A conversation with two engineers. We review your coding exercise, work through a system design problem taken from a real project, and discuss work you are proud of.
- Round 445 minutes
HR interview
Meet HR and a member of the leadership team to talk about motivation, working style, expectations, salary and start date.
- Finish
Offer
We aim to decide within a week of the final round. The whole process usually takes 3–4 weeks.
Apply
Apply for ML Engineer.
Email your resume and a cover letter. It takes about two minutes.
How to apply
Email your resume and a short cover letter, both as PDF, to career@kontextly.ai, with “ML Engineer, Voice & Document AI” in the subject line. We read every application and reply within 5 business days.