Machine Learning PhD While Working in Japan

A TokyoDev article on completing a machine learning PhD while holding a full-time engineering role in Japan reached 39 points and 12 comments on Hacker News.

The post outlines visa constraints, university enrollment rules, and how Japanese employers handle external research commitments. Readers shared concrete timelines and salary figures from their own attempts.

Program Structure and Time Split

Most Japanese universities require PhD candidates to maintain student status for three to five years. Full-time employees typically register as working students and attend evening or weekend seminars.

One commenter reported completing coursework in 18 months while logging 45 hours per week at a Tokyo AI startup. Another noted that lab meetings often occur after 7 p.m., aligning with standard Japanese office hours.

Visa and Enrollment Requirements

International residents on Engineer/Specialist in Humanities visas can enroll part-time without changing status, provided the university issues the necessary documentation. Domestic students face fewer hurdles but still need employer approval for reduced hours during thesis writing.

The article lists specific universities—University of Tokyo, Kyoto University, and Osaka University—that accept working students in machine learning tracks.

Employer Policies Reported on HN

Several commenters described company stances:

• One large electronics firm grants up to 10 hours of paid research leave per month.
• A foreign-owned AI lab permits employees to publish under dual affiliation without salary reduction.
• Two startups required candidates to sign non-compete clauses that blocked thesis topics overlapping with product work.

Output and Publication Data

HN participants shared publication counts. One engineer published three conference papers over four years while maintaining a 9-to-6 schedule. Another reported zero first-author papers after three years due to shifting job priorities.

The discussion highlighted that Japanese conferences such as IEICE and JSAI accept shorter workshop submissions, lowering the barrier compared with NeurIPS or ICML.

Who Should Consider This Path

Engineers already based in Japan with stable visas gain the clearest advantage. Candidates needing frequent international travel or heavy compute resources may find the setup restrictive.

Those seeking rapid career progression in industry should weigh the three-to-five-year commitment against typical promotion cycles at Japanese firms.

Bottom line: The route works best for residents who can align university schedules with existing full-time roles and secure explicit employer support for publication.

The Hacker News thread shows steady interest in Japan-based AI research paths that do not require leaving industry employment.