Generative AI Research and Development Software Engineer

Develop and expand generative-AI products for pharmacy and healthcare. The role focuses on extending the AI assistant that creates medication records from pharmacists’ voices and helping launch several additional AI products. The existing voice medication-record feature processes more than tens of thousands of data items per day and is being expanded as a standard Musubi function.

Responsibilities span model selection, product engineering, and system architecture. The engineer evaluates model trade-offs, designs LLM-based functions such as retrieval-augmented generation, AI agents, document generation, summarization, and structured extraction, and works on prompt and context design and search-accuracy tuning. The role also includes building evaluation infrastructure, operating offline and online evaluations, addressing hallucinations, and designing guardrails for the quality and safety of medical information.

On the product side, the engineer designs and implements front ends with React and TypeScript, as well as APIs and batch processing with TypeScript using Hono and Python. The work includes user experiences specific to generative AI, including streaming, in-progress generation states, and flows for editing and providing feedback on generated results.

The system work covers architecture design and operation using AWS, GCP, and Azure, with infrastructure built through infrastructure as code using CDK. It also includes designing and implementing data pipelines with technologies such as Databricks, and optimizing inference cost, latency, and throughput while maintaining tracing and monitoring for observability. The listed environment includes AWS Athena, ECS Fargate, Aurora, SQS, and Bedrock, Azure OpenAI Service, and GCP Vertex AI. The main LLMs are OpenAI, Gemini, and Claude through Amazon Bedrock.

The engineer works with product managers, designers, and pharmacist domain experts on prototyping, evaluation, specification decisions, feasibility studies, and iterative improvement. The role may also participate in co-creation activities with customers in the medical field. Technical contributions include prompt and model version management, evaluation-dataset maintenance, expanding automated tests, refactoring, library upgrades, and supporting or educating team members in areas of expertise.

The posting does not require immediate expertise across every area. It welcomes candidates with a core strength in one domain who can broaden their scope. The expected level ranges from mid-level to senior-level, with level and responsibilities determined by experience, and independent design decisions are expected at either level. The position belongs to the Data & AI generative-AI research and development team, which combines product development with research activities such as sharing work at domestic academic conferences and submitting to major conferences.

The team consists of one engineering manager, one DS/MLE, five software engineers, one product manager, one pharmacist domain expert, and one designer. Internal counterparts include product management, customer success, SRE, and the DRE team. Development and collaboration tools include GitHub Copilot, Dify, Cursor, Claude Code, Gemini, Devin, NotebookLM, Slack, Google Workspace, Jira, Confluence, Figma, Datadog, and PagerDuty. The selection process is document screening, a casual interview, two or three interviews, an aptitude test, a reference check using back check, and an offer meeting. The listed work location is Tokyo, at 105-0003 Tokyo-to, Minato-ku, Nishi-Shimbashi 2-8-6, Sumitomo Fudosan Hibiya Building 5F.