Based in Japan. I build, observe, and continuously improve AI applications and cloud platforms across RAG systems, CI/CD automation, observability, and secure hybrid infrastructure.
Key domains where I apply production-grade infrastructure, automation, and reliability engineering.
Designing RAG workflows with AWS Bedrock, Dify, OpenSearch, Python, and LangFuse. Built a 10,000+ PDF ingestion and vectorization pipeline, then improved retrieval quality through tracing-driven tuning.
Architecting secure automated delivery with GitHub Actions, Docker, GHCR, Trivy, Cosign, and rollback-aware release flows. Reduced deployment time from 18 minutes to 7 minutes through parallelization and caching.
Implementing Datadog metrics, logs, APM, monitors, and LLM observability. Experienced with GKE production monitoring, LangFuse tracing, and alert rules managed through repeatable infrastructure workflows.
A practical map of how I combine cloud, automation, AI, and observability tools into maintainable systems.
Deploying public services and private lab workloads across cloud and edge networks.
Building repeatable delivery pipelines with container security and controlled service exposure.
Connecting LLM APIs, RAG backends, and personal AI tools into usable application workflows.
Monitoring system health, automating daily workflows, and keeping services easy to operate.
A selection of my personal open-source projects focusing on automation, AI, and reliability.
Serverless RAG API backed by FastAPI and Azure AI Search. Deployed via a modern CI/CD pipeline using GitHub Actions, GHCR, Canary releases, image verification with Cosign, SBOM, and full Datadog APM tracing.
An OpenAI-compatible API proxy hosted on Cloudflare Workers translating calls to Google Vertex AI Gemini. Features SSE streaming support, multiple authentication mechanisms, key rotation, and AI function calling capability.
A lightweight, Cloudflare-native file and text sharing app built with Workers, R2, D1, Vue 3, and TypeScript. Supports extraction codes, anonymous sharing, and resumable uploads as a serverless alternative to FileCodeBox.
My self-hosted playground and multi-cloud environment where I continuously practice system reliability, security, and networking operations.
Representative services from my personal infrastructure. Public demos are linked; admin and storage surfaces stay access-controlled.
Personal LobeChat deployment for multi-model AI usage and prompt testing.
LobeChat / VercelAutomated public transit delay monitor published through GitHub Pages with scheduled data refresh.
Python / GitHub PagesCloudflare-native file and text sharing service backed by Workers, R2, and D1.
Workers / R2 / D1I operate a hybrid multi-cloud and self-hosted infrastructure environment spanning Cloudflare edge, public cloud environments (Azure & GCP), and dedicated home lab servers (Synology NAS).
The dashboard console on the right actively polls my online endpoints to fetch live availability status and round-trip HTTP latency.
Professional credentials validating my expertise in cloud architecture, operations, and AI foundation systems.
Professional
Associate | Oracle Cloud
Associate | Oracle Cloud