Research

Quantitative research powered by data and AI

Alongside enterprise AI applications, we conduct quantitative research on financial markets: building data pipelines, factor models and research dashboards, and making that data directly usable by AI agents. These are our current research projects.

In research

Crypto all-weather portfolio

Portfolio research built on multi-strategy risk parity, spreading risk across market regimes instead of betting on a single direction.

Allocates across strategy legs rather than individual assets, including a trend leg that can go short to fill the hedging gap crypto markets lack
Empirically tests whether classic all-weather assumptions hold in crypto, with quantitative comparisons against established risk-parity products
Validated with strict walk-forward backtesting, keeping leverage low and turnover costs contained

In research

Oil market situation dashboard

A research dashboard combining shipping, inventory, futures-curve and positioning data, designed as an AI-native data layer.

Integrates port and vessel data, official inventory statistics, futures term structure and trader positioning reports
Every value carries its unit, timestamp, methodology and caveats, with as-reported vintages kept to prevent look-ahead bias in backtests
Exposes situation snapshots, historical-analog search and scenario simulation over MCP, so AI agents can reassess the market after breaking events

In internal use

Gold pricing-factor dashboard

Tracks the main drivers of the gold price to help identify the current market regime, used for research and review.

Combines real interest rates, inflation expectations, the broad dollar index, futures positioning, ETF holdings and the futures term structure
Measures gold's deviation from fundamentals with a rolling residual model whose coefficients use only data available at the time
Classifies trending versus range-bound regimes and tracks relative-value signals such as the gold-to-silver ratio

In internal use

Taiwan equity market situation dashboard

Tracks capital flows and positioning in the Taiwan equity market as a daily-updated research dashboard, used for research and review.

Combines institutional net buying, margin balances, institutional futures and options positioning, futures basis and the exchange rate
Official public data first with a third-party backup; every record is tagged with its source, and one source failing never affects the others
Historical rankings use only data available at the time, with automated tests guarding against look-ahead bias

Research principles

No look-ahead: Backtests and factors use only what was knowable at the time, keep as-reported data vintages, and are validated walk-forward.
Record what fails: Research paths that fail their tests are documented as thoroughly as those that pass, so effort isn't repeated and conclusions can be audited.
AI as a first-class consumer: Data is self-describing, carrying its methodology and caveats, and is served to AI agents through standard interfaces such as MCP so large language models can read and reason over it reliably.
Isolated failures: Pipelines use rate limiting, retries and per-source fault isolation, so an outage in one data source never takes down the rest.

This page describes research directions and methods only. It is not investment advice, an offer or a solicitation, and implies no promise of performance. Findings may change as data and methods are updated.

Ready to put AI to work on your workflows?

Tell us about the process you want to improve. We reply within two business days.

Or email us at admin@netlun.io