Hoptroff Timestream and Neural Digital Twins — Since 1988

Timestream and Neural AI that models your world in fine detail, because everything is unique

Precise digital twins of real-world processes and behaviours — to detect anomalies, quantify uncertainty and answer questions conventional models can't. Every application on the right is a real dataset — press play to watch it worked end to end.

Two ways to put your data to work

Learn On The Fly Or Learn The Big Picture

Keep learning from a live stream, or learn once from a dataset — the same neural approach, refined since 1988.

HOPTROFF TIMESTREAM AI · REAL-TIME

Hoptroff Timestream AI

Point the AI at a live stream — markets, sensors, telemetry. It forecasts the next value one step ahead, scores itself honestly on the very next reading, flags anomalies the instant one breaks pattern, and keeps improving online — warm-started from recent history so it's useful from the first minute.

Launch App →

HOPTROFF NEURAL AI · BATCH MODELLING

Hoptroff Neural AI

Turn a spreadsheet into a predictive model. Forecasts and classifications with quantified uncertainty, self-explaining what-if and cross-sections, and one-click export to Excel, VBA and C++.

Launch App →

01 — What we do

Four capabilities that set us apart

Small-data neural modelling

Algorithms refined over three decades extract robust insight from small, noisy, incomplete, inconsistent data. No big-data requirement, no heavy engineering.

Quantified uncertainty

Every twin ships with a twin of uncertainty — a triplet. It tells you what the model predicts and how confident it is, so you can triage.

III

Automated anomaly detection

The triplet defines the expected envelope. Readings outside it — by how much, with what confidence — are flagged and ranked automatically.

Explainable reasoning

Models explain their outputs — quantifying each input's contribution and answering counterfactuals: change this by 10%, what happens?

03 — Why Hoptroff

Built for the real world, not the textbook

Most AI tools assume conditions that rarely exist outside a lab: large, clean, well-structured datasets. The real world is messier — and the gap between what AI promises and what it delivers is usually traced to exactly that mismatch.

Hoptroff was built from first principles to work with data as it is found. Our approach has been refined through real commercial deployments since 1988, across petrochemicals, pharmaceuticals, retail and finance. It doesn't need years of history, and it doesn't break down on gaps, noise or outliers.

The result reaches problems conventional AI cannot — with outputs that are accurate, defensible and explainable, and exportable as C++, VBA or Excel for integration.

1988

Research origins

290+

Worked examples

Triplet

Prediction · uncertainty · anomaly

C++ / XLS

Export for integration

04 — Applications

Where the technology has been applied

Predictive maintenance

Model the expected operating envelope; detect deviations before they become failures; reduce unplanned downtime.

Forensic accounting

Identify anomalous expense claims and records, ranked by confidence. Focus investigation where it matters.

Clinical trials

Quantify drug efficacy and uncertainty from small datasets; decide whether further trials are needed.

Risk analysis

Model contract and investment risk, quantifying return and uncertainty. Spot contracts riskier than they appear.

Time synchronisation

Data centres, telecoms and power grids: tokens-per-watt, edge time-error and localisation from timing data.

Price & demand

Model how sales respond to price, advertising and seasonality; find the profit-maximising price with confidence.

Timestream AI · real-time

Real-time AI that learns as it runs

Where Hoptroff Neural AI learns from a fixed dataset, Hoptroff Timestream AI learns from a live stream. Its flagship is live market data — crypto, FX, indices and commodities — forecasting each instrument one step ahead, flagging anomalies and adapting online, with no retraining runs. Tick the instruments you want and it fetches, models and scores only those.

Live forecasting

Predicts the next value of a stream — markets, sensors, telemetry — one step ahead, scored honestly on the very next reading.

Anomaly flags

Flags a reading the instant it breaks its learned pattern, with a live measure of how surprising it is.

III

Tradeability check

A built-in paper P&L, net of cost, shows whether the forecast edge would hold up against direction hit-rate and buy-and-hold — illustration only, not trading advice.

Online learning

Improves with every reading and adapts as conditions drift — warm-started from recent history, useful from the first minute.

Launch Timestream AI →

Get in touch

Every problem starts with a conversation

We work with a small number of clients at a time, so every engagement gets close, expert attention. Early conversations are always without obligation.

[email protected]