【AI R&D(VLM/LLM)】Full Remote | JLPT N2 level

★AI R&D Engineer (VLM / Multimodal / LLM New Business)｜AI/NLP Company

• Business Level Japanese Required (N2 Level)

◆Flextime＆ Fully Remote Work

◆Start-up Company

◆Own Products/Services

◆Annual Salary：7.5 million ～ 13 million yen.

-------------【About the company】-------------

The company's mission is "to advance humanity by reinventing the mechanisms of value creation" and aims to create an environment where people can focus on creating new value through the use of natural language processing and AI technologies.

AI delivers information directly related to business operations, thereby enhancing the information sensitivity of the organization. In addition, the company offers two SaaS products: "Anews," which encourages the conception of business ideas and their development within the organization, and "Astrategy," a market research service that provides suggestions by structuring necessary information according to the organization and its operations.

Currently, its main target is major manufacturing R&D in Japan, and more than 30% of the Nikkei 225 companies have already adopted the system. In the future, starting with manufacturing R&D, the company plans to expand to domestic enterprise companies, then to Japanese companies as a whole, and even to global companies.

The company completed a Series C financing of 1.1 billion in August 2022.

In February 2024, they were selected for the "GENIAC (Generative AI Accelerator Challenge) Project" promoted by the Ministry of Economy, Trade and Industry and NEDO, and have begun to develop their own LLM with 100 billion parameters!

The company has grown to 100 employees and is looking for more personnel to achieve further growth!

-------------【 Job Description】-------------

【Overview】

The company leverages cutting-edge Generative AI and LLM technology to drive corporate transformation. Its primary strength lies in the technical capability to develop domestic LLMs from scratch. In May 2024, the organization released one of Japan’s largest Japanese-specific LLMs, featuring 100 billion parameters.

While many companies limit themselves to fine-tuning overseas models or utilizing open-source software (OSS), the company designs and builds its models from the ground up. Optimized for the Japanese language and business domains, its performance boasts a standard that is among the highest for domestic models. Furthermore, it has significantly reduced hallucinations, allowing for safe operation under internal management.

Rather than being a mere "user of Generative AI," the company aims to be an "AI creator" that supports Japan's industrial competitiveness through technology. Utilizing this technical prowess, it is expanding its business operations, currently providing AI agents for the manufacturing industry as a SaaS business and an operational AI implementation support platform as a new PaaS venture.

Targeting enterprise companies since its founding, the company has already seen adoption by over 300 firms, representing 30% of the Nikkei 225. Moving forward, it plans to expand from manufacturing R&D to domestic enterprise companies, the Japanese business sector as a whole, and eventually to global corporations.

In October 2024, the company completed a Series D funding round of 4.5 billion yen, bringing its total funding to 8.8 billion yen. With the workforce now reaching 150 members, the organization is recruiting new talent to achieve further growth!

【Current Challenges】

As PoC projects for enterprise clients increase, there is a rapidly growing need to utilize documents that contain not only text but also drawings and diagrams (Visual RAG). While the application of VLM (Vision-Language Model) is key, several challenges remain regarding "accuracy, cost, speed, and reliability" for production-level implementation.

Currently, the company seeks to strengthen its R&D resources to handle everything from executing VLMs-based document structuring PoCs to improving operations and building pipelines for integrating these capabilities into the product as stable features.

【Your Role】

You will lead the research, development, and validation efforts to improve the reliability and functionality of document structuring by leveraging cutting-edge technologies such as VLM. Your role goes beyond simple PoCs; you will be responsible for refining these technologies into "practical, production-ready solutions" built for real-world operations.

■Responsibilities

・Construction and improvement of document structuring pipelines utilizing VLM.

・Enhancement of VLM output accuracy through prompt tuning and the development of validation functions.

・Technical verification and execution of PoCs tailored to client challenges, such as blueprint analysis.

*Scope of changes: Development-related operations.

【Team Composition】

Business Head (CEO): 1

Head of Development: 1

Product Engineers: 7

RAG Engineers: 5

Machine Learning Engineers: 3

Product Manager: 1

Researcher: 1

Customer Success (CS): 1

Business Development (Biz): 3

New Business Planning: 5

【Development Environment】

Languages / Frameworks: Python, TypeScript, Vue.js, Node.js

Containerization: Docker

Infrastructure as Code (IaC): Terraform

Cloud Platforms: AWS, Azure

【Position Highlights】

・Ability to develop and maintain environments that move cutting-edge technologies like VLM and Multimodal AI beyond the laboratory level and into "actual production."

・Opportunity to participate in developing solutions that are not yet established in the market, such as RAG systems specialized for diagrams and charts.

・Access to themes that allow for deep technical exploration, including model operational improvements and pipeline construction, rather than being limited solely to PoC support.

-------------【 Requirements】-------------

【Required】

*Meet all of the following

・Completed a graduate degree in science or engineering.

・Experience in application development or conducting PoCs utilizing LLMs or VLMs.

・Experience in R&D or service development using Python.

・Understanding of the fundamental mechanisms of RAG (Retrieval-Augmented Generation).

・Ability to research and verify unknown technical domains, particularly regarding multimodal models.

【Preferred】

・Knowledge of image processing and Computer Vision (OCR, layout analysis, object detection, etc.).

・Experience in comparative validation of multiple VLM models and designing pipeline architectures.

・Ability to conduct research using academic papers (ArXiv, etc.) and translate findings into practical implementation.

・Experience in translating client business challenges into technical requirements and proposing effective solutions.

【Ideal Applicants】

・Strong intellectual curiosity regarding the latest AI technologies (especially VLM/Multimodal) and a passion for following technology trends.

・Ability to find solutions for highly uncertain technical challenges through iterative experimentation and verification.

・Capability to think beyond "what VLM can do" and focus on "how to make it production-ready" (considering cost, speed, and accuracy).

・Possession of a positive, challenging spirit and the ability to drive development forward in collaboration with a team.

【Others】

・Able to reside in Japan

・Able to communicate fluently in Japanese, including on technical development matters

*Required level: Conversational

--------------------------------------------------

【Working Time 】

Flextime System

■Salary increase (twice a year / May and November)

【About remote work】

■Basically, fully remote work

Residence must be in Japan (no full remote work while residing overseas)

■Flexible remote/working options (no restrictions on the number of times you can come to the office)

Each employee works while using the system according to his/her own circumstances.

There are people who work full remotely from Sendai, Kyoto, Osaka, etc., people who enjoy working while on vacation, foreign nationals who return to their home countries for a few weeks to work, and people who mainly come to work, making for a diverse work style.

【Welfare】

■Customer facing allowance (flat monthly amount of 20,000 yen *included in monthly salary)

Free use for self-improvement and customer value improvement

(purchase of books/attendance at seminars/equipment, etc.)

■Lump-sum allowance for preparation for remote work (flat rate of 50,000 yen in the first paycheck)

Provided as an expense for preparation for remote work

【Others】

■Periodic health checkups (all paid by company)

■Influenza vaccination

■PC of your choice provided

■Free drinks such as water/coffee/carbonated drinks/alcohol provided at the office

■Office Glico installed

【Education and Skill Development】

■Onboarding for 5 days after employment is available

■Monthly study day system (for own skill improvement and learning purpose)

【Internal Communication】

■Weekly company-wide online meeting

■Monthly social gathering (offline, some online)

■Semi-annual kick-off (hybrid of online and offline meetings)

【Holidays and Vacations】

■123 days off per year

■Saturdays, Sundays and national holidays (2 days off per week)

■Year-end and New Year vacations

■Paid leave (granted at the time of employment *Number of days granted depends on the month of employment)

■Prenatal leave

■Childcare leave