AI-Focused Software Engineering — Tokyo

LumoraTech

A Tokyo engineering team building AI systems that run in production.

Engineered in Tokyo

We turn hard problems into systems that hold up under real traffic — retrieval pipelines, agent architectures, and the full-stack products and data platforms around them.

Our work Get in touch

Projects shipped

AI systems live

Years of engineering

Scroll

02 — WORK

Selected Work

Production systems we designed, built, and shipped end to end.

View all

OpenClaw-style Personal AI Agent

Autonomous AI executive assistant that connects to Slack, Calendar, GitHub, and email — handling your daily workflow through natural language.

LangGraphClaude 4 SonnetGemini 2.5 ProNext.js 15

Production RAG Enterprise Search

Instant-answer engine over company documents with 92% accuracy, hybrid retrieval, source citations, and eval-gated deployments.

Next.js 15FastAPIpgvectorColBERT

AI Trading Terminal

Full Stack

Professional-grade trading dashboard with live candles, 25+ technical indicators, AI-powered signals, paper trading, and backtesting.

Next.js 15FastAPIWebSocketsRedis

AI Business Workflow Automator

End-to-end sales automation — lead qualification, CRM updates, demo scheduling, and follow-up sequences handling 85% of SDR workflows.

LangGraphClaude 4 SonnetHubSpot APICalendly

View all projects

03 — STACK

Our Technology Stack

AI & Machine Learning

Python

PyTorch

LangGraph

LangChain

PydanticAI

OpenAI

Anthropic

Hugging Face

RAG

LLMs

Computer Vision

Vector DBs

Fine-tuning

Agents

Full-Stack Engineering

Next.js

React

TypeScript

Node.js

PostgreSQL

Docker

AWS

Drizzle

Redis

tRPC

Tailwind CSS

Vercel

04 — SIGNALS

Writing & Experiments

Notes from our engineering practice, and the experiments we run in the lab.

Journal

View all

AI EngineeringFebruary 20, 2026

Building Production RAG Systems That Actually Work

Most RAG tutorials show toy examples. Here's what it really takes — hybrid retrieval, eval-gated CI/CD, chunking strategies that survive messy documents, and the metrics that matter in production.

12m

AI EngineeringJanuary 15, 2026

Designing Multi-Agent Systems with LangGraph

How to architect agent supervision graphs, handle inter-agent communication, implement human-in-the-loop approval, and avoid the deadlock traps that plague multi-agent workflows.

10m