Tech Lead at Japan AI · Tokyo, Japan
Building the plumbing behind enterprise AI
I make LLM agents fast, reliable, and accounted for, down to the last token.
I'm a backend and platform engineer with 8+ years of experience building high-throughput, cloud-native systems. These days I work on the platform behind Japan AI's enterprise AI products, which span chat, tool-using agents, speech, and Studio for custom AI apps. My focus is agent runtime infrastructure, usage metering and billing, LLM gateway integration, and evals.
- Japan AI, Tech Lead (current). I work on the platform and agent infrastructure behind Japan AI's enterprise AI products, including metering, quota and entitlement systems for LLM usage.
- PayPay Corporation, Senior Backend Engineer. I built backend systems for a payments platform serving 65M+ users, contributed to an ML-assisted fraud detection pipeline (XGBoost on SageMaker), and built a production RAG pipeline with GPT-4 and OpenSearch k-NN at sub-100ms p99.
- Earlier roles spanned enterprise, fintech, data engineering and mobility. Highlights include an NLP search service handling 15k req/s on AWS EKS, a 50% cut in P95 latency, a 20% reduction in cloud spend, and a Rust BigQuery-to-Snowflake SQL translator.
- LLM and agent platforms, covering agent runtimes, LLM gateways, RAG, and evals.
- Usage metering, quotas and billing pipelines.
- Event-driven microservices on Kafka, gRPC and Kubernetes.
- Data engineering and high-performance services in Rust.
Languages: Python, Rust, Java, Kotlin, TypeScript Backend: FastAPI, Spring Boot, gRPC, Kafka, PostgreSQL, Redis, OpenSearch AI/ML: OpenAI and Anthropic APIs, LangChain, LiteLLM, vector search, PyTorch, scikit-learn, XGBoost Infrastructure: GCP, AWS, Kubernetes, Docker, Terraform, Helm