AI/NLP Engineer

This role covers the core AI technology (RAG, VLM, and LLMOps) behind Stockmark's SAT (Stockmark A Technology) product, currently in a product-market-fit phase. Responsibilities include improving the RAG pipeline and designing new algorithms, improving the accuracy of Agentic RAG components, and designing and driving LLMOps, including evaluation infrastructure, automation, and scaling. The role also covers multimodal document understanding: researching and evaluating Vision-Language Models (VLM) for multimodal document analysis, designing and implementing a business-document analysis engine and its API, and exploring technical approaches to document-analysis problems that general-purpose LLMs cannot solve. The engineer connects research results to the product, working with machine learning and product engineers on model requirements and API/data-structure design, and improving Deep Research, knowledge-graph, and in-house LLM inference applications. As part of technical leadership during the PMF phase, the role involves working with product managers to investigate customer technical challenges, technical evaluation for enterprise PoCs, and feeding results back into the product and algorithms. The development environment uses Python, TypeScript, Vue.js, and Node.js, with Docker, Terraform, and AWS/Azure cloud infrastructure. Employment is full-time, with flexible remote work available to residents of Japan.