// EDGION · TECH NATION ENDORSED

UK‑engineered intelligence. built for your idea.

// decision platforms · trading signals · climate models

// for teams with a problem worth solving

> scroll to read
// THE METHOD · ENGAGED

we model uncertainty.
you make the call.

// the engagement ends. the model doesn't.

// the client re-runs it. on new data. next year.

geo N51°30'26" W00°07'40"
wind 12kt @ 240°
vis 3.2km · ceil 230m
press 1014 hPa · rising
link edgion · reading · · 42ms
[ MACHINE READY ] edgion@reading:~ > scroll to inspect
// 01 · WHAT WE BUILD

the machine //
for your decision.

// we work across domains.

// what stays constant is the engine.

Trading signals
Quant models, AI and NWP weather signals, systematic strategies, backtested and live. Built to perform in production, priced at real friction.
Climate & hazard
Physics-based ensemble simulations of future climate. Quantified shifts in typhoon, flood, and wind hazard. Grounded in physical dynamics, validated against observation.
Market intelligence
Entry analysis, competitive positioning, quantified opportunity sizing. Built to re-run when the market shifts.
Decision platforms
Custom risk engines, scoring systems, model-driven decision pipelines. The infrastructure you own and operate.
// 02 · METHOD

three rules //
we never break.

// three practices that make the work repeatable.

// all of them load-bearing.

RULE.001

engineer‑built.

the deliverable runs in production.

pipelines, models, platforms your team operates.

re-run it whenever the question comes back.

RULE.002

calibrated.

every output ships with its confidence bounds.

a narrow interval: act. a wide one: gather more data first.

the uncertainty is the information.

RULE.003

first‑principles.

we derive models from physics and economics.

every assumption is explicit and auditable.

built to work where the data runs out.

// 03 · EVIDENCE

two cases //
verifiable on call.

// names withheld on request.

// numbers are real.

CASE.001 · BESPOKE UNDERWRITING

How do you price loans against an engineering asset that hasn’t been built yet?

Built: end-to-end risk-to-credit platform for a UK clean energy lender. Engineering performance risk formally coupled to credit risk. Five-stage workflow from device telemetry to loan portfolio. Production-grade Python.
→ pricing loans against unbuilt assets. systematic, auditable, client-owned.
CASE.002 · WEATHER ALPHA

When a weather model updates its forecast before delivery, does that revision carry information the power market hasn’t priced yet?

Built: quant-ML fusion pipeline. Signal: NWP and AI weather model revision delta, the gap between consecutive forecast updates for the same delivery window. Walk-forward validation on years of European power data. Auditable backtest, client-owned, deployment-ready.
→ combined Sharpe 2.42 · p = 0.002 · win rate 53%
// 04 · START

tell us //
the decision.

// we read every brief personally · reply within one business day.

officeDavidson House, The Forbury, Reading RG1 3EU, United Kingdom