Backend Engineer, ML Platform

This backend engineer will own the application layer of a credit assessment microservice used by digital banking and fintech products across the group. The service combines aggregated data calculated by Databricks, model endpoints on Amazon SageMaker, and internal credit calculation logic to provide credit information to product teams. The role is responsible for converting probabilistic and uncertain machine-learning outputs into deterministic API behavior that financial products can depend on.

Responsibilities include designing and implementing a high-performance, low-latency microservice for a scale of tens of millions of users, including service boundaries and data ownership. The engineer will design OpenAPI schemas, error-code conventions, versioning, backward compatibility, and processes for converting business requirements from stakeholders into API contracts. The role also covers SageMaker endpoint client design, latency-budget allocation, timeouts, retries, circuit breakers, and fallback behavior.

The position includes implementing credit-domain logic, covering boundary and abnormal cases, ensuring idempotency, and designing reproducibility and audit trails that link input data, model versions, and business logic. Performance work includes reading aggregated data from S3, managing data freshness and switching consistency, caching, and efficient asynchronous or parallel model calls. Quality responsibilities include unit, contract, and load-test strategy, regression verification when models are replaced, structured logging, traceability, and application instrumentation based on service-level indicators.

The engineer will lead architecture selection, code reviews, development-process improvement, and standardization. The role works autonomously in a small Scrum team and collaborates closely with SRE and infrastructure engineers responsible for infrastructure, security, monitoring, and CI/CD, and with machine-learning engineers responsible for model development. The service is being developed for the planned digital-bank launch in fiscal 2027 and must support high traffic and high availability.

Required experience includes approximately five or more years of software engineering practice, including designing, developing, and operating Web APIs or microservices with Python or similar technologies; experience building high-traffic, low-latency applications and tuning performance; leading failure and contract design for external-system integrations; establishing automated testing processes; and developing applications on AWS with CI/CD and container-based environments. Candidates must be able to discuss technical decisions with SRE, infrastructure, and machine-learning engineers and lead specification development with stakeholders.

Preferred experience includes MLOps and model-integration knowledge, S3, Glue, or Databricks data-platform understanding, Terraform configuration, financial, credit, or payments-domain experience, security requirements such as mTLS, OAuth, or JWT, technical-lead or Scrum Master experience, and AI-assisted development. The named stack includes Python, FastAPI, Amazon API Gateway, AWS ECS Fargate, S3, CloudWatch, Amazon SageMaker, GitHub Actions, Docker, Terraform, Databricks, AWS Glue, Claude Code, Slack, and Notion.

The position is full-time in Minato, Tokyo. It uses a hybrid work style with two office days per week required in principle and three or more office days recommended, subject to company and business conditions. Core working hours are 9:30–18:30 with a 60-minute break, under a discretionary work system where applicable. Annual compensation is JPY 7,008,000–11,004,000, with possible semiannual high-performance bonuses. The trial period is three months from the start date. Business-level Japanese and basic business-level English equivalent to TOEIC 700 or above are required. Benefits include social insurance, housing support, paid leave, summer and winter leave, year-end holidays, defined-contribution pension, employee stock ownership, book subsidies, health checks, and conference support.