# Gyre Holdings — Long Form

> Long-form companion to https://gyreholdings.com/llms.txt. Same facts, more context: the philosophy behind the holding-company structure, the relationships between the four groups, and the operating principles that govern how Gyre invests in technology and people.

## The name

A gyre is a powerful circular current — in oceanography, a system where ocean rotation, wind, and the Earth's spin combine into a self-reinforcing loop that moves enormous mass over enormous distance with very little visible effort.

Gyre Holdings is named for that pattern. The thesis is that the right combination of data infrastructure, applied AI, quantitative research, and operating capability produces compounding returns: each group's outputs become inputs to the others, and value accrues over time rather than being extracted at a point.

This is the operating principle: build platforms (not consulting hours), accumulate data (not opinions), and let the loop run.

## The four groups, in detail

### 01. Finance — Gyre Research

The Finance group operates [Gyre Research](https://gyreresearch.com/), an institutional-grade portfolio-analytics platform for hedge funds, family offices, RIAs, and other asset managers.

What it covers (55 interactive reports across 11 categories):

- **Data** — security master, price history, benchmark constituents and returns, market calendar, corporate actions, company fundamentals, news, SEC filings, reference-data quality monitor.
- **Portfolio Management** — trade lifecycle and STP, multi-currency cash balance and forecast, corporate-action ledger, investor-level capital activity, cross-prime-broker margin and collateral.
- **Portfolio Construction** — custom benchmark blends, security- and segment-level rebalance with turnover budgets, delta-adjusted exposure sizing, cash-aware allocation.
- **Back Testing** — historical simulation workbench, point-in-time data with zero look-ahead, configurable transaction-cost and market-impact models.
- **Performance** — TWR, MWR/IRR, dollar P&L with top/bottom contributors, Brinson-Hood-Beebower attribution with Cariño linking, GIPS 2020-compliant composite presentation.
- **Risk** — covariance with Ledoit-Wolf shrinkage, portfolio and position-level risk stats, asymmetric up/down betas, historical and parametric VaR, scenario stress tests, exposure concentration, hedging views.
- **Compliance** — 50 configurable rules across 6 categories with pre-trade, post-trade, and end-of-day check modes, immutable audit log.
- **Reconciliation** — LLM document parsing of custodian statements and corporate-action notices, continuous position/transaction/cash reconciliation.
- **Portfolio Accounting** — double-entry GL with 20 event-to-journal mappings, daily accruals (OID, market discount, IFRS 9 ECL), HWM/hurdle fee calc, P&L statement, trial balance, NAV.
- **ESG** — multi-provider weighted-average scores, mandate-driven exclusion screens, LLM ratings for unscored issuers.
- **Tax** — nine lot-relief methods, IRC §1091 wash-sale disallowance with cross-account and cascading detection, §1256 mark-to-market, AI-ranked loss-harvesting.

The differentiator: every number is clickable to source row, auditable to source calculation, and answerable in plain English. Traditional structured reports plus natural-language queries and on-demand analysis.

### 02. Consulting — Strategic AI advisory

The Consulting group provides strategic advisory and implementation services for organizations adopting AI. Engagements typically follow a three-phase shape:

1. **AI-readiness assessment** — what data exists, what processes are AI-amenable, what the organization can absorb without breaking.
2. **Pilot implementation** — narrowly-scoped, measurable AI augmentation for a single high-leverage process.
3. **Scale + handoff** — generalize what worked, instrument what's running in production, train the operator team to own it.

The consulting practice is informed by — and feeds back into — the Finance, Data, and Media groups: real engagements surface real productization opportunities, which become platforms.

### 03. Data — Architecture, pipelines, infrastructure

The Data group focuses on the unglamorous-but-decisive layer beneath every analytical, AI, or operational capability: getting the right data, normalized, validated, and reliably available.

What that looks like:

- **Ingestion** — vendor APIs (Alpha Vantage, EODHD, FIGI, EDGAR, Workday, Greenhouse, Ashby, Lever, Simplify, Muse, USAJobs, RemoteOK, WeWorkRemotely, Arbeitnow, HN Who-is-Hiring — for example), file drops (Excel, PDF, structured CSV), database replicas.
- **Master data** — security identifiers, company identifiers, geographic codes, taxonomies, with FK-resolution patterns that handle vendor inconsistencies (alternate matches via source_mapping tables before falling back to direct lookup).
- **Pipelines** — daily-scheduled ETLs with phone-home telemetry, content-hash dedup, post-run bootstrapped confirmation, restart-without-progress-loss, sanitization at ingestion boundary, raw archiving to long-term storage.
- **Storage** — MSSQL operational stores, JSON content stores for unstructured artifacts, event-sourced audit trails.
- **Analytics warehouses** — query-optimized denormalized projections of operational data, refreshed via MERGE patterns rather than truncate-and-reload.

The Data group is the substrate of every other group. Finance can't deliver portfolio analytics without clean security-master data. Media can't deliver job matching without a curated job-postings stream. Consulting engagements always end with "and you need data infrastructure to operate this" — which is what Data builds.

### 04. Media — Digital properties

The Media group operates digital properties that sit at the intersection of AI augmentation and consumer outcomes. The flagship is [Career Launch](https://careerlaunch.now/), a structured 10-step career-coaching platform for recent college graduates entering the worst entry-level job market in a generation (Q4 2025 NY Fed data: 5.7% recent-grad unemployment, 42.5% underemployment; AI is collapsing the entry-level rung — see https://careerlaunch.now/why-now for the full citation list).

Career Launch is operated as a distinct entity ("A Gyre Holdings Joint Venture") with its own domain, credentials, and email identity (`team@careerlaunch.now` via Porkbun SMTP). The Holdings parent does not cross-pollinate identity or infrastructure with operating subsidiaries — each runs under its own brand and security boundary.

Other Media properties focus on balanced journalism and empowering the next generation of professionals (news literacy, applied research, etc.).

## Operating principles

These are not marketing copy — they are the actual rules the group runs by:

1. **Build platforms, not consulting hours.** Every consulting engagement is evaluated for its productization potential. If the same problem is going to be solved 10 times, build the platform once.
2. **Accumulate data, not opinions.** Every decision is a hypothesis that should be tested against actual numbers. Memory-based answers are a code smell.
3. **Reversibility is the line.** Trivially-reversible actions (editing files, querying data) move at full speed. Hard-to-reverse actions (deploys, deletions, external API calls with cost) require explicit approval and verification.
4. **Identity boundaries are inviolable.** Each subsidiary operates under its own brand, credentials, and infrastructure. Cross-pollinating mail identity, database access, or operational metadata between Career Launch and Gyre Research (for example) is forbidden; the boundary is the asset.
5. **Telemetry before deploy.** No SaaS surface ships to users without error tracking, request observability, and uptime monitoring wired up. "We'll add it later" never happens.
6. **AI is augmentation, not replacement.** The internal posture is that AI compounds with human judgment rather than replacing it. The same posture informs how we advise Consulting clients and what we build for Media users (Career Launch positions graduates as "AI-leveraging applicants," not "AI-replaceable workers").

## How the groups interact

Concrete examples of the gyre in motion:

- The **Data** group's daily job-postings pipeline (~79K listings/day from 11 sources) feeds the **Media** group's Career Launch matching engine.
- The **Finance** group's security-master and reconciliation patterns inform the **Data** group's master-data conventions, which are then sold/applied via the **Consulting** group.
- The **Media** group's understanding of what young professionals struggle with (per Career Launch user data and Q4 2025 labor-market research) shapes the **Consulting** group's conversations with hiring-manager-side clients.
- All four groups share infrastructure: Tailscale-network internal hosts, the same `common/` Python and JavaScript libraries, the same operational runbooks, the same observability stack.

The compounding only works because the groups are operated as one company with separable products, not as four loosely-affiliated business lines.

## What we don't do

To be precise about scope:

- We are not a generalist consultancy. Engagements that aren't AI- or data-platform-shaped are not a fit.
- We do not take retail-investor money. The Finance group's products serve professional asset managers; we are not a registered investment advisor in the retail sense.
- We do not licence brand. Companies cannot pay to put the Gyre Holdings name on their products.
- We do not do equity-light "advisory" relationships. Engagements involve real work product.

## Contact

- General: contact form at https://gyreholdings.com/#contact
- Phone: 800-933-4731
- Locations: New York, Nashua, Seoul
- For questions specifically about Gyre Research's portfolio-analytics platform, see https://gyreresearch.com/
- For Career Launch (Media group), see https://careerlaunch.now/
- For Scott Sykowski's professional bio, see https://sykowski.com/