Company Overview: We plan and operate "Pairs", a dating and marriage matching app for people who are genuinely looking for a partner. The service launched in Japan in 2012, a Taiwan version followed in 2013, and a Korean version was released in 2025. Cumulative registrations exceed 27 million (as of April 2026), and many people use the service. In addition, more than 900,000 people have met their irreplaceable partners through Pairs.
Job Summary: The AI Team is responsible for developing and operating AI-powered products for Pairs, including moderation to protect Trust & Safety and recommendations. We run LLM and machine learning applications that handle text and image review, judgment of violation reports, fraud detection, and more, on infrastructure built around Kubernetes across AWS and Google Cloud. The MLOps Engineer in this role joins the AI Team and embeds within each AI product team (a way of working close to Embedded SRE), taking responsibility for balancing product reliability with development speed. You will have end-to-end ownership of the Kubernetes cluster itself, shared components on the cluster, online inference APIs, asynchronous workers, batch processing, and continuous learning pipelines. Working with the SRE, security, and privacy teams and product teams, you will build the foundation that supports our AI products. The main mission is not model research and development itself, but creating a state where AI products run stably in production and can be improved quickly and safely. We also expect you to actively use coding agents and AI agents to reduce operational toil through systems. In addition, we expect you to feed the knowledge gained from your work and implementation back into the harnesses for AI agents in the form of documentation, lint, and tests, and to cultivate an environment where people and agents can develop safely and quickly together.
What You'll Do: Upgrading Kubernetes (Amazon EKS) clusters, managing node groups and autoscaling, and maintaining and improving cluster components managed with GitOps (Argo CD / Helm / Kustomize) Redesigning observability, monitoring, and alerting for AI applications (improving metrics, logs, and traces, and building actionable alerts and Runbooks), and automating first-level alert investigation using AI agents Defining and operating SLIs / SLOs for AI applications in collaboration with the SRE team (setting targets based on failure impact for each use case, burn rate alerts, and building triage flows across teams) Designing and operating a system to continuously update dependency libraries, Python runtimes, base images, and security patches for the many Python applications that make up our AI products (automated update bots, lockfile management, and CI-based verification) Leading migrations associated with updates to LLM models and SDKs (such as generational migrations of Claude on Amazon Bedrock, GPT models, and Gemini models on Gemini Enterprise Agent Platform), along with quality regression checks and phased releases Promoting security and data protection for AI infrastructure in collaboration with the security and privacy teams (vulnerability response and permission management in light of supply chain attacks targeting OSS packages, container images, and CI/CD, as well as attacks made more sophisticated and automated by AI, and handling training data and logs in line with data retention rules) Operating and improving continuous learning pipelines, training environments, and dataset preparation environments using Argo Workflows and Amazon SageMaker Operating and improving the environment on Gemini Enterprise Agent Platform (formerly Vertex AI) used for training and batch processing in the recommendation domain Operating online inference APIs, queue-driven asynchronous workers, and batch applications running on Kubernetes (capacity planning, cost optimization, incident response, and improving CI/CD and release flows) Reducing daily development and operational toil using coding agents and AI agents Building a system that feeds knowledge gained from daily work and implementation back into harnesses that AI agents can use for reference and verification (documentation, lint, tests, CI checks, etc.), and continuously growing them
Basic Qualifications: At least 3 years of hands-on experience implementing, releasing, and operating commercial applications Experience developing production applications (including package management, type checking, and testing) Experience creating and operating container images used in production environments (writing Dockerfiles, introducing middleware, etc.) Experience upgrading Kubernetes clusters and operating GitOps with Argo CD / Helm / Kustomize Deep understanding of either AWS or Google Cloud, with experience building and operating production infrastructure Experience with incident response and system improvement using metrics, logs, and traces Experience designing and operating batch and asynchronous processing that accounts for idempotency and retries Experience developing with coding agents and a strong motivation to use AI Ability to understand how the code you write behaves in the runtime environment, to face failures head-on, and to resolve them through your own hands-on work Ability to prioritize based on value to products and business, rather than treating building infrastructure as an end in itself Ability to reconcile technical ideals with reality and move things forward while building agreement with stakeholders Language skills: Japanese (business level or above), English (conversational level or above)
Preferred Qualifications: Experience operating machine learning model training pipelines or inference services in production Experience developing or operating applications that incorporate LLMs (such as Claude or Gemini on Amazon Bedrock), or experience with model migrations Experience operating workflow engines such as Argo Workflows or Airflow Experience designing and introducing SLIs / SLOs, or leading observability improvements using tools such as Datadog Experience with IaC using tools such as Terraform Experience introducing and operating automation for dependency updates and security patching (such as Renovate or Dependabot) Experience with software supply chain security measures (such as vulnerability management for dependencies and container images, protecting build and release paths) or cloud permission design Experience designing and operating systems that handle data containing personal information in line with retention and purpose-of-use rules Experience improving CI/CD and release engineering using tools such as GitHub Actions Experience practicing operational automation with AI agents (such as supporting incident investigation and reducing toil) Experience maintaining instruction files for AI agents (such as AGENTS.md), lint, and tests, and continuously improving the quality of agent work Experience using or operating ML platforms such as Amazon SageMaker or Gemini Enterprise Agent Platform (formerly Vertex AI) Experience building and operating data pipelines using BigQuery or Amazon Athena Experience with Policy as Code using tools such as OPA Gatekeeper or Conftest Experience learning new technologies on your own and working hands-on across a broad range of areas without being limited to a single domain
Tools we use: Programming languages: Python, Go, Shell Applications / ML: FastAPI, Uvicorn, LightGBM, scikit-learn, pandas, Polars, PyTorch LLM: Amazon Bedrock (Claude, GPT), Google Gen AI SDK (Gemini) Container platform: Amazon EKS, Argo CD, Argo Workflows, KEDA, External Secrets Operator, OPA Gatekeeper, Cluster Autoscaler, Helm, Kustomize AWS: Amazon SageMaker, Amazon SQS, Amazon S3, Amazon Athena, AWS Glue, Amazon Aurora (MySQL compatible), Amazon DynamoDB, etc. Google Cloud: BigQuery, Gemini Enterprise Agent Platform (formerly Vertex AI), etc. IaC: Terraform Observability: Datadog (APM, Logs, Monitors, Synthetics), Fluent Bit Security: Wiz, OPA Gatekeeper, Conftest Development: Git, GitHub, GitHub Actions, Docker, uv, Ruff, mypy, pytest, Renovate / Dependabot, Jira Cloud, Confluence, Slack AI tools: Claude Code Enterprise, Codex / ChatGPT Enterprise, Cursor Enterprise, Gemini Enterprise
#Eureka
#LI-AW2