This position leads the implementation and operation of credit assessment models for Money Forward Kessai (MFK) and the Digital Bank project.
You will migrate models from the PoC phase to a bank-grade production environment, establish continuous quality management through MLOps, and drive implementation of fairness and transparency based on the principles of Responsible AI.
Background
Initial validation of the credit assessment models is largely complete. As the next phase, we are accelerating both “integration into banking systems” and “building a strict operational setup.” As a hands-on leader, you will be expected to drive solutions to the following challenges:
- Migration to production: Smoothly migrate from the verification environment (Databricks) to the production environment (SageMaker), and build a robust CI/CD pipeline that removes uncertainty.
- Bank-grade quality assurance: Implement XAI (e.g., SHAP, counterfactuals) to explain lending decisions, and detect/mitigate model bias to ensure fairness.
- Efficient processing of large-scale data: Optimize training and inference cycles by introducing distributed learning/processing (e.g., pandas UDF / Spark) leveraging the characteristics of accounting transaction data.
Technology Stack
- Platform: AWS (SageMaker, Lambda, ECS, S3, etc.), Databricks
- Data: Python (FastAPI, etc.), SQL, Apache Airflow / Step Functions
- DevOps: Terraform, GitHub Actions, CodePipeline
- Communication: Slack, Notion
Responsibilities
- Design and implement ML workflows
- Build an end-to-end ML pipeline integrating Databricks and Amazon SageMaker.
- Implement and enhance credit assessment models
- Implement scalable model training using large datasets (e.g., Parquet).
- Introduce fairness and explainability (XAI)
- Implement bias evaluation using statistical methods (e.g., statistical parity) and algorithms to visualize decision rationales.
- Model operation and monitoring
- Continuously monitor performance degradation and environmental changes (concept drift), and operate retraining workflows.
Requirements
- Experience operating ML models in production
- Practical experience across the full lifecycle from development to deployment and post-release monitoring.
- Experience in data processing and development with Python
- Practical use of ML libraries (pandas, scikit-learn, etc.) and backend frameworks.
- Technical leadership experience
- Experience conducting design reviews and setting technical direction in a small team.
- Fundamental data engineering knowledge
- Experience using RDBMS/SQL and building data pipelines on cloud platforms (e.g., AWS).
- Business level Japanese (equivalent to JLPT N2 or above). Please note that the interviews in the selection process will be conducted in Japanese.
- Basic business level English (equivalent to TOEIC 700 or above)
Nice to haves
While not specifically required, tell us if you have any of the following.
- Knowledge of AI fairness and explainability
- Understanding of fairness metrics and use of tools such as SageMaker Clarify.
- Experience with distributed processing at scale
- Performance tuning experience using Apache Spark, pandas UDF, etc.
- Knowledge of the financial domain
- Experience developing systems in banking/credit operations, or in environments compliant with security standards such as FISC.
Compensation
¥6,408,000 ~ ¥11,004,000 annually.
