We build for
Precision
A young, ambitious product team in Seoul. We are commissioned to plan and build systems that move a lot of data.
The hard part is rarely the screen. It is the data underneath it.
Ingestion that survives a dropped connection. Figures that reproduce a year later. A computation somebody can audit. Teams bring us in when that layer is the risk, and the interface still has to be good.
We build the systems that decisions run on.
Eight areas, from the pipeline to the interface. One team holds all of them.
Real-time data pipelines
Streaming ingestion, normalisation, fault recovery, back-pressure
Third-party integration at scale
Rate limits, spec drift, dropped connections, replays that never double-count
Metric and indicator engineering
Definitions, thresholds, regime logic, validation against history
Blockchain and on-chain indexing
Node and provider ingestion, address labelling, flow computation
AI features grounded in your data
Summarisation and reading, with the basis and the limits stated
Web and mobile product
Next.js and React, iOS and Android, design systems for dense data
Cloud infrastructure and operations
AWS architecture, CI/CD, monitoring, alerting, cost control
Data integrity and reconciliation
Settlement, cross-checking, correctness under load and after replay
Five things we hold to.
Measured, not assumed
Every figure we ship carries its calculation, its sources and the history of its definition. If a number cannot be reproduced, it does not go out.
Built to be verified
Recomputation and lineage go into the first schema rather than being retrofitted. The same input returns the same output a year later, and we can show why.
We operate what we build
We run systems in production ourselves, so the design accounts for the 3am page, the reprocessing job and the audit request. The handover includes the runbook.
One scope, end to end
Ingestion, computation and interface are held by one team. Nothing is lost at a handover that never has to happen.
Plain language over jargon
A figure nobody can read is not a result. We state what a number means, and where it stops being reliable.
A crypto market data platform, built end to end.
A consumer platform for market data, taken from first commit to general release. Four exchange integrations over REST and WebSocket, normalised into a single model, with rankings computed from real trading activity rather than self-reported figures.
A framework of 24 indicators spanning derivatives, spot market structure, on-chain flows, macro and sentiment. Each one carries its calculation, its sources and the thresholds that decide which regime it is reporting, plus an AI layer that reads them back in plain language.
Shipped on web, iOS and Android in six languages, with the monitoring and the operating documentation needed to run it after handover.
We are hiring.
We hire on a rolling basis. A young, ambitious team in Gangnam, Seoul.
Tell us the data and the goal.
An unfinished specification is fine. Business, investment and hiring all reach one address.