— About

Operators, not consultants.

A research team built for the next crisis — not the last one. Below: the people, the procurement answers, and the path from first email to ninety-day pilot.

Team makeup.

Built where it has to work. Not where it gets pitched.

01 · Research

Research & modelling.

Ph.D.-trained financial engineers. Backgrounds at sovereign wealth funds and quantitative buy-side desks. Published research across regime detection and financial NLP.

02 · Engineering

Quantitative engineering.

Veterans of top-tier proprietary trading desks. Production architects behind LLM-augmented factor models. On the execution and market-data path, latency budget is measured in microseconds.

03 · Relations

Institutional relations.

Sourced from global private wealth desks and sovereign mandate teams. Onboarding for APAC and EMEA family offices and sovereign funds runs through this seat.

Operating principles

Built where the tail has to be read, not where models get pitched. Ph.D.-led research. Veteran quantitative engineering. Onboarding sourced from global private wealth desks. Every output carries auditable provenance — built for regulators, not for slide decks.

— Frequently Asked

Questions from institutional desks.

For procurement, security, or technical onboarding queries not covered here, we respond within one business day.

What distinguishes mindsight-galaxy's LLM from conventional VaR models?+

Conventional parametric VaR assumes normality and stationarity — assumptions that fail precisely when they matter most. Our system overlays a generative reasoning layer that ingests policymaker discourse, supply chain disclosures, and counterparty filings to dynamically reweight the tail. The output is a regime-conditional VaR, not a historical average.

Which institutions is this designed for?+

Sovereign wealth funds, family offices managing $250M+ AUM, quantitative hedge funds, pension funds, and corporate treasury desks with material multi-asset exposure. We do not serve retail.

How is data privacy and model security handled?+

Two deployment modes: (1) On-premise vault with air-gapped inference on your hardware, or (2) dedicated single-tenant cloud with SOC 2 Type II controls and encrypted-at-rest model weights. Portfolio data never traverses public inference endpoints.

What is the training corpus?+

Licensed datasets from major financial archives (1985–present), full-text central bank communications across 17 jurisdictions, SEC and equivalent regulatory filings, and proprietary annotated crisis-event corpora. No client portfolio data is used for training.

How does pricing work?+

Annual institutional license with three tiers based on AUM band, inference throughput, and deployment mode. Pilots typically run 90 days with full SLA. Contact our institutional relations team for indicative pricing.

What integration paths exist?+

REST API, FIX gateway adapters, Bloomberg Terminal companion plugin, and direct integration with Aladdin, Charles River, and Eze Investment Suite. Average time-to-first-signal in a new deployment: 11 business days.

Is model output auditable for compliance?+

Yes. Every output carries a deterministic provenance trace: source documents, intermediate inference steps, and regime classification confidence. Designed to satisfy FINRA, SFC, FCA, and MAS audit requirements.

How do we start a pilot?+

Three-stage onboarding: (1) 30-minute discovery call to align scope, (2) signed mutual NDA + technical workshop, (3) 90-day pilot with measurable KPIs and a defined off-ramp. We do not lock institutions into multi-year contracts before they have validated edge.

— Institutional Briefing

Let's discuss what your tail looks like.

A confidential 45-minute briefing with our institutional relations lead. Walk away with a regime-mapped summary of your current exposure.

Email

horizon.lin722@taiwan-culture.com

Drop a line or use the form. Either reaches us.

Confidential. One business day response.