Live in production Forex · Gold · Real-time

Trading software that reacts as fast as the market.

TensorStar is where I build real-time tools for people who watch price for a living. The first is Quadrant — a desk that streams six timeframes at once, draws its signals straight onto the chart, and replays any session bar by bar.

A personal project by one engineer — not a company, and not a commercial service. Access is private, by invitation only.

Product
Quadrant
Status
Running 24/5
Timeframes
1m → 4h
Transport
WebSocket
Built by
Truong Mai Huy
quadrant.tensorstar.com
The Quadrant desk: three synchronised XAUUSD charts across 1m, 5m and 15m, each with its own RSI pane, annotated price levels and a live readout.
◆ XAUUSD · three timeframes live ◆ Signals drawn on the chart ◆ Actual screenshot, not a mockup
01Product

Quadrant. The whole picture, on one screen.

Four synchronised charts, live signals and your journal in one place — so you stop switching tabs in the middle of a trade.

TIMEFRAMES

1m → 4h

Six of them, up to four on screen, synced by time or bar count.

6
ALERTS

Straight to Telegram

Three detection engines fire the moment a setup confirms.

3engines
WORKSPACES

One live dataset

Charts, calendar, journal and capital plan — all off the same stream.

4
REPLAY

Study the past without hindsight

Rewind any session and step bar by bar. Indicators recompute exactly as they would have at that moment — nothing from the future leaks in.

CALENDAR

News on the chart, not in another tab

High-impact events are drawn onto the timeline, so a release never catches you mid-position.

Quadrant is an analysis tool, not advice. Its detectors mark patterns on a chart; they are not recommendations, predictions, or any indication of future results. Nothing here or in the app is financial advice. Trading forex and gold carries a high risk of loss.
Sign in · invite only →
02Engineering

Built for speed, not for show.

Every layer is ours — ingestion, indicator maths, realtime transport, deployment, and the last pixel on screen.

01

Streaming, not polling

Prices, forming bars and indicator values arrive over a live WebSocket. The chart updates in place — it never blanks out to refresh.

02

Signals computed server-side

RSI trendlines, double tops/bottoms and session bands are detected once on the backend and shared by the chart and the alert bot, so what you see is exactly what fires.

03

Replay without hindsight

The replay engine slices data at the cursor before indicators run, so a backtest can never read a bar that hadn't closed yet.

04

Operated, not just deployed

Containerised, TLS-terminated, cache-warmed and monitored on my own infrastructure — with a runbook for every failure mode I've hit.

PythonFastAPIWebSocket ReactTypeScriptRedis DockerCaddy
03About

Built by one engineer.

I'm Truong Mai Huy, an AI engineer in Da Nang, Vietnam, building real-time computer-vision and machine-learning systems since 2019.

I also trade forex and gold. Quadrant is the desk I wanted and couldn't find, so I designed, built and run it myself — every layer, from the data pipeline to the deployment.

04Contact

Let's build something that holds up.

Questions about Quadrant, a collaboration, or custom work — I'd like to hear from you.