How to Apply to 株式会社PKSHA Technology

9 min read Last updated April 20, 2026 51 open positions
Blake Crosley
Blake Crosley Researched using publicly available information, hiring data, and ATS analysis.

Key Takeaways

  • PKSHA Technology is a publicly listed Tokyo-headquartered enterprise AI company (TYO:3993) with roughly 600 employees and roots in University of Tokyo machine learning research.
  • Applications go through HRMOS, the BizReach ATS — submit Japanese-format 履歴書 and 職務経歴書 as PDFs and complete every free-text field in the candidate portal.
  • Daily working language at HQ is Japanese; JLPT N2 is the practical minimum for engineering roles, N1 expected for senior IC, management, and customer-facing positions.
  • Some research roles and global expansion roles accept strong English-primary candidates, but these are the exception rather than the rule.
  • Compensation is competitive within Japan but typically below the Tokyo offices of US-headquartered AI labs (Anthropic JP, Google JP, Lily Asia) — the trade-off is deeper enterprise problem exposure and a shorter Hongo commute.
  • Interviews are technically rigorous and culturally formal; prepare to defend your past project decisions in detail and to discuss long-term career intent honestly.
  • Through 2024-2025 the company integrated LLMs across the PKSHA suite, opening more roles around RAG pipelines, agentic systems, and voice-LLM integration.

About 株式会社PKSHA Technology

PKSHA Technology Inc. (株式会社PKSHA Technology) is a publicly listed Japanese AI and algorithm company headquartered in the Hongo neighborhood of Bunkyo Ward, Tokyo — a stone's throw from the University of Tokyo campus where the company's intellectual roots lie. Founded in 2012 by Katsuya Uenoyama, a Tokyo University Ph.D. researcher in machine learning, PKSHA began as an algorithm-licensing shop and evolved into one of Japan's most credible domestic enterprise AI players. The company listed on the Tokyo Stock Exchange under ticker 3993 in 2017 and today employs approximately 600 people across its core entity and operating subsidiaries. PKSHA's flagship product portfolio sits under the PKSHA suite: PKSHA Communication (a chatbot and conversational AI platform widely deployed at Japanese banks, insurers, and telcos), PKSHA Voicebot (voice-driven customer service automation), PKSHA Chatbot (FAQ and self-service automation), and PKSHA Workplace (an enterprise productivity AI layer that integrates with Microsoft Teams and other internal tools). Through 2024 and 2025 the company aggressively integrated large language models into every product line, retrofitting legacy rule-based bots with retrieval-augmented generation pipelines and shipping new agentic features for back-office automation. The strategic shift moved PKSHA from a domain-specific algorithm vendor toward a horizontal applied-AI platform that competes directly with both domestic SI players and the enterprise arms of foreign cloud providers. Uenoyama still leads the company as CEO and remains its public face, often appearing at NIKKEI AI summits, University of Tokyo industry-academia events, and earnings calls where he frames PKSHA's mission as 'a future where humans and software co-evolve.' The customer base skews heavily toward the Japanese enterprise heartland — megabanks (Mitsubishi UFJ, Mizuho, SMBC have all been PKSHA customers at various points), property and casualty insurers, large retailers, telecommunications carriers, and government agencies. This customer composition anchors the day-to-day work in long sales cycles, careful on-premise security reviews, and the painstaking craft of deploying AI inside organizations that move carefully and weigh institutional risk above iteration speed. For candidates, PKSHA occupies a specific and distinctive niche in the Tokyo AI labor market: it is more research-oriented than a typical Japanese SaaS startup, more product-focused than a pure research lab, and more domestically anchored than the Tokyo offices of Google, Anthropic, OpenAI, or Lily Asia. Compensation is competitive by Japanese standards but typically below the Tokyo offices of US-headquartered AI firms. The trade-off is depth of problem space, real customer impact at Japan Inc., a strong academic-engineering culture rooted in University of Tokyo lineage, and a shorter commute from the Hongo academic corridor.

Application Process

  1. 1
    Visit the PKSHA Technology recruit site at recruit

    Visit the PKSHA Technology recruit site at recruit.pkshatech.com and review both the 新卒 (new graduate) and 中途 (mid-career) tracks — they run on separate timelines and use different evaluation rubrics.

  2. 2
    Identify the role family that fits: Algorithm Engineer, Software Engineer, Produ

    Identify the role family that fits: Algorithm Engineer, Software Engineer, Product Manager, Solution Architect, Sales, or Corporate. Research roles often have additional publication or thesis requirements listed in the job description.

  3. 3
    Create your candidate profile in HRMOS (the BizReach-owned ATS PKSHA uses)

    Create your candidate profile in HRMOS (the BizReach-owned ATS PKSHA uses). You will need a Japanese-format 履歴書 (rireki-sho) and 職務経歴書 (shokumu-keireki-sho) for mid-career applications; new grads typically submit an ES (entry sheet) instead.

  4. 4
    Submit your application through the HRMOS portal

    Submit your application through the HRMOS portal. Expect an automated acknowledgment within minutes and a substantive recruiter response within 5–10 business days; LLM and research roles can take longer due to hiring committee review.

  5. 5
    Complete the document screening (書類選考) stage

    Complete the document screening (書類選考) stage. PKSHA recruiters look hard at GitHub, Kaggle rankings, and arXiv preprints for engineering candidates, so make those links easy to find on your resume.

  6. 6
    Move through 2–4 interview rounds: a recruiter screen, one or two technical inte

    Move through 2–4 interview rounds: a recruiter screen, one or two technical interviews (algorithm coding, system design, or research deep-dive depending on role), and a final round with a senior engineer or director. Some product roles include a take-home case.

  7. 7
    If selected, you will receive a verbal offer (内々定) followed by a written offer (

    If selected, you will receive a verbal offer (内々定) followed by a written offer (内定) with compensation, signing bonus where applicable, and start date. New grads typically start in April; mid-career hires negotiate flexible start dates.

  8. 8
    Sign the offer and complete onboarding paperwork through HRMOS

    Sign the offer and complete onboarding paperwork through HRMOS. Pre-employment background checks are standard for finance-sector-facing roles.

Resume Tips for 株式会社PKSHA Technology

recommended

Submit a Japanese-format 履歴書 with photo and a separate 職務経歴書 for mid-career appl

Submit a Japanese-format 履歴書 with photo and a separate 職務経歴書 for mid-career applications. PKSHA recruiters expect this convention; an English-only US-style resume will read as not-yet-localized for Japan and may cost you the document screen.

recommended

If you are applying from overseas or for an English-friendly research role, subm

If you are applying from overseas or for an English-friendly research role, submit both Japanese and English versions. State your JLPT level explicitly (N1, N2, N3) and your business Japanese comfort level — vague language about 'conversational Japanese' is read as N3 or below.

recommended

Quantify algorithmic impact in production terms: 'Reduced inference latency from

Quantify algorithmic impact in production terms: 'Reduced inference latency from 280ms to 45ms across 4M monthly requests,' 'Improved chatbot first-contact resolution from 62% to 78% across three megabank deployments.' PKSHA's customer base values measurable enterprise outcomes.

recommended

List specific ML frameworks, MLOps tooling, and LLM stacks: PyTorch, JAX, vLLM,

List specific ML frameworks, MLOps tooling, and LLM stacks: PyTorch, JAX, vLLM, LangChain, LlamaIndex, Triton, ONNX, Vertex AI, Azure OpenAI Service. Generic 'machine learning' phrasing is invisible at the screening stage.

recommended

Highlight University of Tokyo, Tokyo Institute of Technology, Kyoto University,

Highlight University of Tokyo, Tokyo Institute of Technology, Kyoto University, or comparable degrees if you hold one — PKSHA's hiring culture has a strong academic-pedigree signal, especially for research-track roles. International degrees from top-tier programs are equally valued.

recommended

For mid-career candidates, name your previous employers' industry segments (mega

For mid-career candidates, name your previous employers' industry segments (megabank, P&C insurer, SaaS, consumer internet) so reviewers can map your domain experience to PKSHA's customer base.

recommended

Include open-source contributions, Kaggle medals, or NeurIPS/ICML/ACL/EMNLP publ

Include open-source contributions, Kaggle medals, or NeurIPS/ICML/ACL/EMNLP publications prominently. PKSHA's algorithm engineering team treats these as primary evidence of capability rather than supplementary credentials.

recommended

Keep total length to 2 pages for the 職務経歴書 — Japanese hiring norms penalize spra

Keep total length to 2 pages for the 職務経歴書 — Japanese hiring norms penalize sprawl. Use the standard reverse-chronological format with explicit start and end dates in YYYY/MM.

ATS System: HRMOS by BizReach

PKSHA Technology runs its applicant tracking on HRMOS, a Japanese ATS built by BizReach (now part of Visional). HRMOS is the dominant ATS for mid-to-large Japanese tech employers and is tightly integrated with the BizReach scout database, meaning your application data and any prior BizReach scout history can be visible to recruiters. The system supports both 新卒 (new graduate) and 中途 (mid-career) workflows on separate pipelines, and it handles the document review, scheduling, and offer stages end-to-end. The candidate-facing portal is primarily Japanese, with limited English coverage for global hires.

  • Fill out every HRMOS profile field — including self-introduction (自己PR) and motivation (志望動機) free-text boxes — even if your resume covers the same ground. Recruiter dashboards surface these fields prominently and blank ones read as low engagement.
  • Upload PDFs (not Word docs) for both 履歴書 and 職務経歴書. HRMOS preserves PDF formatting reliably; .docx files can render inconsistently in the recruiter view.
  • If you have a BizReach profile, make sure it is current and consistent with your HRMOS application — recruiters often cross-reference the two and inconsistencies (especially around dates and titles) raise concerns.
  • Use the HRMOS messaging thread for all scheduling and follow-up rather than email. Recruiters track candidate communication inside the ATS, and out-of-band emails get lost.
  • Set HRMOS notifications to email or LINE so you do not miss interview scheduling windows — PKSHA's recruiting ops team often sends scheduling links with 48-hour response expectations.

Interview Culture

PKSHA's interview process is rigorous, technical, and conducted primarily in Japanese.

For algorithm and research roles, expect to walk through past projects in detail — interviewers will ask probing questions about your model architecture choices, hyperparameter selection, evaluation methodology, and how you handled production failure modes. The bar for technical depth is closer to a top-tier research lab than to a typical Japanese SaaS company; surface-level answers do not pass. Engineers should prepare for whiteboard or shared-document coding sessions covering data structures, dynamic programming, and at least one ML-flavored design problem (recommendation system, retrieval pipeline, or LLM agent design). Culture fit is evaluated through behavioral questions about how you handle long enterprise sales cycles, on-premise customer constraints, and the slower iteration cadence that comes with serving Japanese megacorp customers. PKSHA values intellectual humility, written communication discipline (engineers are expected to write detailed design docs in Japanese), and a willingness to take customer-site meetings in person. Candidates who frame themselves purely as 'move fast and break things' founders often misread the room. The tone in interviews is courteous, formal, and quietly demanding — interviewers will not raise their voices, but they will keep pressing on weak answers. Final-round interviews frequently include a meeting with a director or executive, sometimes Uenoyama himself for senior research hires. These conversations probe long-term career intent (PKSHA hires for tenure, not for two-year sprints) and your view on where applied AI is going in Japan specifically. Come prepared with a thoughtful, evidence-based answer.

What 株式会社PKSHA Technology Looks For

  • Demonstrated ability to ship ML or LLM systems into production at scale, not just prototype in notebooks
  • Strong fundamentals in algorithms, probability, and linear algebra — Tokyo University, Tokyo Tech, and equivalent academic backgrounds are heavily represented
  • Comfort working in Japanese for daily standups, design docs, and customer-facing meetings (typically JLPT N2 minimum, N1 for senior IC and management roles)
  • Track record of working with enterprise customers under tight security, compliance, or on-premise deployment constraints
  • Curiosity about applied research — engineers are expected to read papers, attend NLP/ML conferences, and bring ideas back into the product
  • Patience and judgment for the long sales and integration cycles typical of Japanese megabank, insurer, and government customers
  • Written communication discipline: design docs, postmortems, and customer-facing technical writeups in Japanese
  • Long-term career intent — PKSHA hires for tenure and invests heavily in onboarding, so two-year-mercenary signals work against candidates

Frequently Asked Questions

Is PKSHA Technology a startup or an established company?
PKSHA Technology has been publicly listed on the Tokyo Stock Exchange (ticker 3993) since 2017 and employs around 600 people across its core entity and operating subsidiaries. It operates with the discipline of a public company — quarterly disclosures, investor relations, formal governance — while retaining a research-oriented engineering culture inherited from its University of Tokyo origins. It is not a startup in the venture-backed sense, but it is also not a legacy SI.
Do I need to speak Japanese to work at PKSHA?
For nearly all roles at the Hongo headquarters, yes. The day-to-day working language is Japanese — standups, design documents, code review comments, and customer meetings all happen in Japanese. JLPT N2 is the practical minimum for engineering positions, with N1 expected for senior individual contributors, managers, and any role that touches customers. A small number of pure research roles and roles supporting global expansion will consider English-primary candidates, but these are exceptions and you should confirm explicitly during the recruiter screen.
How does PKSHA compare to Anthropic JP, Google JP, or other foreign AI labs hiring in Tokyo?
The foreign AI labs typically pay more in cash and equity, work in English by default, and focus on frontier research or global product development. PKSHA pays competitively for the Japanese market but generally below US-HQ labs, works in Japanese, and focuses on applied AI for Japanese enterprise customers — banks, insurers, retailers, and government. The trade-off is depth of domain exposure to Japan Inc. that you will not get at a foreign lab, and a more stable, long-tenure career path. Different candidates rationally choose differently.
What ATS does PKSHA Technology use?
PKSHA uses HRMOS, an applicant tracking system built by BizReach (Visional). HRMOS is the dominant ATS among mid-to-large Japanese tech employers. The candidate portal is primarily in Japanese, integrates with the BizReach scout database, and handles document review, interview scheduling, and offer management end-to-end. Submit your 履歴書 and 職務経歴書 as PDFs through the HRMOS application form.
Does PKSHA hire new graduates (新卒) and mid-career (中途) candidates differently?
Yes. The 新卒 track follows the traditional Japanese hiring calendar with applications opening roughly a year before the April start date, ES submissions, group interviews, and a unified onboarding cohort. The 中途 track runs year-round with rolling applications, individualized interview schedules, and flexible start dates. Both tracks use HRMOS but evaluate candidates against different rubrics. New grads are evaluated on potential, academic credentials, and culture fit; mid-career hires are evaluated on demonstrated production track record.
Where is PKSHA's office and is remote work available?
PKSHA's headquarters is in the Hongo area of Bunkyo Ward, Tokyo, near the University of Tokyo main campus. The company has historically operated a hybrid model with a strong in-office expectation, particularly for engineering and customer-facing roles where on-premise customer visits and team collaboration matter. Fully remote work is uncommon at PKSHA; expect to be in Tokyo and in the office multiple days per week.
What is the interview process like for an algorithm engineer role?
Typically 3–4 rounds: a recruiter screen, one or two technical interviews covering coding (algorithms and data structures), system or ML system design, and a deep-dive into your past projects, then a final interview with a senior engineer or director. Expect probing questions about model architecture decisions, evaluation methodology, and production failure modes. Bring concrete examples with numbers and be ready to discuss them in Japanese.
Does PKSHA work with large language models?
Yes — extensively. Through 2024 and 2025 PKSHA integrated LLMs into every flagship product: PKSHA Chatbot, PKSHA Voicebot, PKSHA Communication, and PKSHA Workplace. The work spans retrieval-augmented generation pipelines for FAQ automation, voice-LLM integration for call center automation, and agentic systems for back-office workflows at enterprise customers. Roles touching RAG, fine-tuning, evaluation, and agent orchestration are actively hiring.
Who founded PKSHA Technology and who runs it now?
PKSHA was founded in 2012 by Katsuya Uenoyama, a University of Tokyo Ph.D. researcher in machine learning. Uenoyama remains CEO and is the public face of the company at NIKKEI AI summits, University of Tokyo industry-academia events, and earnings calls. The leadership bench includes co-founders and senior executives drawn from both academic AI backgrounds and Japanese enterprise software companies.
What kind of customers does PKSHA serve?
PKSHA's customer base is anchored in Japan Inc.: megabanks (Mitsubishi UFJ, Mizuho, SMBC at various points), property and casualty insurers, large retailers, telecommunications companies, and government agencies. This means engineering and solutions work involves long sales cycles, careful security and compliance reviews, on-premise or private-cloud deployments, and a meticulous customer-success culture. Candidates who thrive at PKSHA enjoy this depth of integration; those who prefer fast-iteration consumer SaaS typically do not.

Open Positions

株式会社PKSHA Technology currently has 51 open positions.

Check Your Resume Before Applying → View 51 open positions at 株式会社PKSHA Technology

Related Resources

Similar Companies

Related Articles

Sources

  1. —
  2. —
  3. —
  4. —
  5. —
  6. —
  7. —
  8. —