Machine Learning Engineer

This machine learning engineer will structure general documents containing text and tables into usable customer data. The work combines in-house VLMs, other LLM/VLMs, image processing, and PDF analysis according to the problem.

Responsibilities include improving table-extraction and layout-analysis algorithms, including preprocessing and postprocessing; designing and implementing document-structuring logic; expanding the product application of the in-house VLM; optimizing routing accuracy and cost across in-house and external LLMs and VLMs; and designing fallback mechanisms for model or API failures and accuracy degradation.

The role also generalizes capabilities proven in specific customer use cases and integrates them into the product. It includes designing generalization logic for non-standard formats such as customer Excel data, extracting statistics from customer-data analysis, and building a data flywheel for accuracy improvement. This includes designing evaluation metrics and preparing evaluation and training datasets. The engineer works with the product team and Customer Success to identify and improve accuracy issues arising in the field. The stated scope may also include development and operation of company services and products, related work, and other duties designated by the company.

The position is part of the PaaS Unit. The listed team consists of one CEO serving as business lead, one development lead, seven product engineers, five RAG engineers, three machine learning engineers, one product manager, one researcher, one Customer Success member, three Biz members, and five new-business-planning members. The company context includes an AI-agent SaaS product for manufacturing, a platform for implementing business AI, and in-house development of Japanese LLM and VLM technologies.

The named development environment includes Python, TypeScript, Vue.js, Node.js, Docker, Terraform, AWS, and Azure.

The stated work location is the Tokyo headquarters or the employee’s home or another place suitable for remote work, with no change in scope. Full remote work is available and is generally the standard arrangement. The employee must reside in Japan; full remote work while living overseas is not available. Residents within the company’s commuting area of Tokyo and neighboring prefectures may choose remote work or office attendance without a frequency limit, while residents elsewhere in Japan may voluntarily attend the headquarters once per month.