• to navigate
  • to select
  • to close
  • System Development · AI Infrastructure Strategy · Implementation Support

    Modern Tech
    No Fuss
    Systems That Last

    With 11 years on GCP plus AWS, Azure, Cloudflare, and Akamai, I specialize in cloud system development. Backend, API, infrastructure automation, data platform × LLM integration — designed, built, and delivered end-to-end by one engineer.

    One Person, Like a Team

    loFT LLC is a technology-focused company run solely by Koushuu Matsubara. Founded as a sole proprietorship in 2014 and incorporated in 2019. From requirements gathering through design, implementation, infrastructure provisioning, and operations — all handled in-house with no outsourcing. This prevents the common issue in large organizations where design intent is lost in operations, maintaining consistency while enabling continuous improvement. I have supported diverse companies’ cloud systems in a hands-on, one-on-one partnership model. Especially with startups, I’ve driven design, development, and operations with fast decision-making and unwavering design intent.

    Founded / Location

    2014 (incorporated 2019) · Musashino, Tokyo

    Specialization

    Cloud infra, AI/LLM platforms, full-stack development, DevOps

    2026 Focus — LLM Infrastructure R&D

    Experiencing the potential of open-source models in local environments, I’m focusing on LLM infrastructure research and development.

    Research Infrastructure

    Built an on-premise high-performance computing environment with AMD EPYC 9175F and RTX PRO 6000 (96GB×2). Designing, building, and validating quantization, multi-model inference, AI data pipelines, and modern data stack infrastructure.

    System & Tool Development

    Focused on building systems and tools for local LLM use cases — inference optimization, MCP tools (Rust), data pipeline integration — advancing local development infrastructure.

    Contract Development

    Taking on projects in system/service design, full-stack development, data platforms, multi-LLM integration, RAG systems, inquiry and internal knowledge chatbots (open-source LLMs), application development for AI model productization and cloud infrastructure, and GPU analytics infrastructure design and build.

    Strengths

    Because I handle everything from design to implementation, decisions are fast and design intent stays consistent.

    11 Years of GCP Cloud Architecture

    Building on Google Cloud Platform since 2015 — GKE / Cloud Run / Firebase for core infrastructure. Multi-cloud and edge configurations including AWS, Azure, Cloudflare, and Akamai with container-based microservice design.

    Full-Stack Execution That Moves Products Forward

    From UI design (Figma) to frontend (React/Vue/SolidJS) and backend (Go/Python/Rust) — all end-to-end. Since technology selection and implementation are the same person, spec changes are tracked quickly.

    Technology Selection Driven by Requirements

    Working across Python/Go/Rust/TypeScript/Dart and many more languages, selecting the optimal language and architecture for each requirement.

    Reproducible Operations with DevOps / IaC

    Designing deployment and configuration automation on Kubernetes / Podman containers. Observability built in from the start with Prometheus + Grafana + Loki. Most services I’ve supported are microservice-based, with operations designed to fit each team.

    LLM/AI Infrastructure Design & Validation (On-Prem GPU)

    Continuously validating open-source large language models and datasets from Hugging Face on our own EPYC 9175F + RTX PRO 6000 (96GB×2) on-premise GPU infrastructure.

    Data & Analytics Platforms (Modern Data Stack)

    Design and operations experience with PostgreSQL / AlloyDB / Firestore. In-house development of data pipeline and analytics infrastructure with Apache Iceberg + dbt + Trino + Dagster.

    Tech Stack

    Key technologies used over the years.

    Languages

    • Python
    • Go
    • Rust
    • TypeScript
    • JavaScript
    • Dart
    • Kotlin
    • C#
    • awk
    • Shell Script
    • SQL
    • Swift

    Frontend / UI

    • React
    • Vue
    • SolidJS
    • Next.js
    • Nuxt
    • Alpine.js
    • Tailwind CSS
    • Bootstrap

    Mobile / Desktop

    • Flutter (iOS/Android/Web)
    • Tauri (iOS/Android/Web)
    • iOS (Swift)
    • Android (Kotlin)

    Cloud

    • GCP (11 years)
    • AWS
    • Azure
    • Cloudflare
    • Akamai (Linode)
    • Vultr

    Containers / OS

    • k8s
    • Istio
    • Docker
    • Podman
    • Container-Optimized OS
    • Ubuntu
    • Debian
    • CoreOS
    • Alpine Linux

    Databases / DWH

    • PostgreSQL
    • MySQL
    • Firestore
    • MongoDB
    • AlloyDB
    • DuckDB
    • PostGIS
    • Postgres-vector

    Data Platforms

    • Apache Iceberg
    • dbt
    • Trino
    • Dagster
    • Nessie
    • Lightdash

    AI / LLM / ML

    • vLLM
    • llama.cpp
    • ik_llama.cpp
    • Ollama
    • PyTorch
    • TensorFlow
    • LoRA
    • Qdrant
    • ONNX Runtime

    GIS / Maps

    • MapLibre
    • Mapbox
    • rust-martin

    Observability

    • Prometheus
    • Grafana
    • Loki
    • Promtail
    • Various exporters

    Dev Tools / Ops

    • Zed
    • VSCode
    • Figma
    • Gitea
    • Git
    • Obsidian
    • Cloud Build
    • Cloud Run Jobs
    • Cloud Scheduler
    • Terraform

    Storage / Backup

    • GCS
    • S3
    • Cloudflare R2
    • MinIO