HNWI Intelligence Engine

High-net-worth individuals leave structured public footprints — a new VCC in Singapore, a PSC filing in the UK, a Golden Visa grant in Malta — weeks or months before any of it surfaces in Pitchbook, Preqin, or a wealth-report PDF. The HNWI Intelligence Engine reads those footprints as they hit the registry, scores them, maps the relationships around them, and turns each one into a briefing and outreach draft. The operator gets a two-month head start on every other firm chasing the same capital.

The problem it solves

The ultra-high-net-worth segment is uniquely unreachable through digital marketing. Meta and Google ads don't land. Cold email blasts get filtered. Gatekeepers — family offices, advisors, lawyers — stand in front of every inbox. The people behind the wealth respond to one thing: a peer who shows up with specific, verifiable knowledge of their corporate structure before the rest of the market has noticed.

The gap is structural. Capital moves through registry filings — new family offices, holding companies, SPVs, citizenship-by-investment grants, freehold property acquisitions. Commercial databases (Pitchbook at $25,000+/year, Preqin at $15,000+/year, Refinitiv at $20,000+/year, Orbis at $10,000+/year) ingest that data with a 2-8 week lag, in a format optimized for fund analysts rather than for outreach intelligence. A solo operator or a 5-person advisory firm can't justify those subscriptions, and even those who can subscribe pay for delayed, generic data with no relationship mapping and no scoring tuned to their own thesis. The choice today is six-figure spend on stale information or weeks of manual OSINT that doesn't scale past a handful of targets per month.

Who needs this most

• Solo founders and partners selling into the UHNW segment — anyone shipping a $50K-500K product (a trading platform, a fund-administration tool, a structuring service) that cannot be sold through paid digital channels and must reach the principal directly, with credibility, before a competitor does.
• Boutique capital-introduction firms (3-15 people) placing capital into funds and direct deals, who need to surface deployment-ready HNWIs from registry data instead of relying on the same recycled lists every other intro firm is working from.
• Family-office advisors and structuring lawyers in 2-3 jurisdictions who need to know within the same week when a target client moves their structure — a new SPV, a fresh director appointment, a Singapore VCC formation — so the follow-up call happens before another advisor has had it.

The moment this hurts: any week when the operator has a $50K-500K offering, knows the buyer profile exists, and watches the calendar slip by with no credible way to land in front of the right principal before someone else does.

The solution — in plain terms

The engine reads the registries the same way a fund analyst would, but does it continuously and at scale. Every six hours it pulls new incorporations, director changes, and beneficial-ownership filings from UK Companies House. Daily runs cover Singapore BizFile+, Hong Kong ICRIS, Estonia's e-Business Register, and the Irish CRO. The Malta and Thai Royal Gazettes feed in citizenship and visa grants. Dubai's Land Department contributes freehold property purchases. Each fresh filing becomes a Signal in a structured database.

From there the engine does three things that a manual analyst cannot. It links the entities — a new holding company is connected to the director who set it up, who is connected to two other companies they sit on, who is connected to a partner who appears as PSC on a fourth entity. The web of relationships becomes an interactive graph rather than a stack of PDFs. It scores each individual on a 0-100 Liquidity Strength Score combining signal type, recency, structural complexity, jurisdiction quality, and cross-border presence — a single number that says "this person is deploying capital right now." And it generates a per-individual briefing note plus a personalized outreach draft, ready to send, ready to defend, sourced back to the original registry URL.

The operator's weekly rhythm collapses to thirty minutes on Monday reviewing the dashboard, an hour on Tuesday on the top three scored individuals, an hour on Wednesday refining the AI drafts, thirty minutes on Thursday sending. Four hours a week instead of forty.

Value delivered — what you get

• A 2-8 week head start on Pitchbook, Preqin, and every advisor refreshing the same commercial dataset — the engine ingests the filing the day it goes public; commercial feeds catch up a month or two later.
• Replaces $15,000-25,000/year in single-source commercial subscriptions — the engine is self-hosted, runs on one VPS, has no per-seat licensing and no per-jurisdiction surcharge.
• Turns outreach from cold-pitch roulette into peer-to-peer observation — every draft references a specific entity, a specific filing, and a specific structural insight, sourced back to the registry URL. The response rates on "I noticed your new Singapore VCC overlaps with your UK holdings" are not in the same universe as generic vendor email.
• Maps the corporate structure around a target before the first conversation — directors, shareholders, PSCs, linked SPVs and assets — surfaced as an interactive force-directed graph, exportable to Maltego CSV for advanced analysis.
• Flags distress signals as a leading indicator, not a postscript — insolvency notices, winding-up petitions, late filings, and forced share transfers feed the same scoring engine, so risk and opportunity show up on the same dashboard.
• Compresses the solo-operator outreach week from forty hours to four — dashboard review, briefing generation, draft refinement, and send-tracking all live in one admin panel.
• Keeps every signal sourced and auditable — every record links back to the original registry URL, so any claim made in outreach or in a compliance review can be defended against the underlying public filing.

Where it delivers outsized value

• Solo software founders selling high-ticket products into the UHNW segment — trading platforms, fund tooling, structuring services where the buyer is a principal at a family office and there is no advertising channel that reaches them.
• Boutique capital-introduction and fund-placement firms working across 2-4 jurisdictions, where surfacing fresh deployment-ready prospects each month is the actual revenue activity.
• Compliance and AML-adjacent teams running pre-screening on prospective clients, where beneficial-ownership cross-references across multiple registries are the daily workflow and the alternative is a $50K+ enterprise feed.
• Wealth advisory and family-office firms operating in two or three jurisdictions where knowing the same week that a client has restructured is the difference between a renewed retainer and a quiet departure.

The common thread: situations where the value of one well-timed conversation with the right principal is six figures or more, and where the existing tools (commercial databases, generic CRMs, manual OSINT) cannot deliver that conversation at the moment it matters.

Distinctive features — why this over the alternatives

• Liquidity Strength Score — a single 0-100 number that means something — six weighted components (signal type, recency, structural complexity, jurisdiction quality, cross-border presence, distress penalties) collapse a thicket of filings into one prioritization signal. Thresholds are tunable from the UI so the operator can calibrate against their own response data.
• Three-tier scraping with honest fallback — BeautifulSoup first (free and fast), Playwright headless when JS rendering is required, Playwright with a visible browser when anti-bot defenses bite. CAPTCHA solving runs pytesseract OCR with an Ollama LLaVA vision fallback. No scraping vendor in the loop.
• Local-first AI with cloud as fallback, not as default — entity classification, HTML extraction, and CAPTCHA solving run against a local Ollama instance at no per-token cost. Briefing notes and outreach drafts can route to Anthropic, Mistral, or OpenAI when the writing quality justifies the spend. The router decides per task type.
• JSONB-flexible schema that adapts to new signal types — registries publish new filing categories on their own schedule; the scoring engine reads new signal types from a config table rather than a code release, so adding a Maltese citizenship grant or a Dubai DLD transaction is a database row, not a migration.
• Interactive force-directed graph plus Maltego export — the in-app graph (react-force-graph) handles same-day analysis; Maltego CSV export hands the data to seasoned OSINT analysts who want to take it deeper.
• Compliance-first sourcing posture — every record carries its source URL, every dataset is public-record only, no auth bypass, no paywall jumping, and the data retention rules are baked into the scoring engine. Defensible to a legal review.

Under the hood — built to last

Python 3.12 with FastAPI for the API surface, SQLAlchemy 2.0 async on PostgreSQL 16 with JSONB flex fields, Celery on Redis for the scraping and scoring queues, React 18 with TypeScript and Tailwind for the admin panel. Seven containers stand the whole platform up on a single VPS or an on-premise box, with an optional eighth GPU-backed container for the local Ollama runtime. Nothing depends on a SaaS subscription to function — every layer can be operated offline, behind a firewall, or in a sovereign-data environment. Twenty-eight markdown documents (whitepaper, architecture, data model, scoring model, scraping playbook, AI integration, deployment) ship in the repo as canonical specification, so the system is operable by someone who never spoke to the original author.

Current maturity

The platform is mid-build and operational on a core slice. The full Docker Compose stack stands up cleanly. The UK Companies House scraper is live — search, company detail, officers, and PSC endpoints with the three-tier fallback in place — and is the reference implementation for the four Tier 1 scrapers still in build (Singapore BizFile+, Hong Kong ICRIS, Estonia Ariregister, Ireland CRO). All eleven frontend pages are real implementations, not stubs: dashboard, jurisdictions, entities, individuals, signals, graph, scores, scrapers, AI tools, export, settings. Live scrape progress streams to the UI via three-second polling. The scoring engine schema and Liquidity Strength Score model are specified in full and partially wired; AI-router infrastructure is in place with provider stubs ready for the Ollama, Anthropic, Mistral, and OpenAI chains. Approximately 9,400 lines of code across backend and frontend, with ~5,900 in Python and ~3,500 in TypeScript, as of the most recent session on 2026-03-08. Active development, not yet a publicly deployed product.

Roadmap — what's next

The immediate path forward is the remaining four Tier 1 scrapers (Singapore, Hong Kong, Estonia, Ireland) plus the OpenCorporates cross-reference layer — once those are live, the engine moves from a single-jurisdiction proof-of-concept to a real cross-border intelligence surface, and the Liquidity Strength Score starts behaving as designed because cross-border signals are a core component of the model. Tier 2 jurisdictions (Luxembourg LBR, Dubai DLD, Malta Gazette PDF parsing) follow, opening the engine to the Gulf and Mediterranean wealth corridors. Wiring the full AI integration (briefing notes, outreach drafts, the Ollama-first router) is the milestone that converts the platform from data collection into intelligence packaging.

Longer-arc, the engine moves from a personal operator tool into a sellable intelligence product — per-jurisdiction subscription tiers priced against the $15K-25K commercial floor, an API tier for FinTech and compliance integrations, and a white-label option for family-office advisory firms and capital-introduction houses that want to attach the underlying intelligence layer to their own branded offering. The roadmap document tracks each phase against explicit exit criteria.

Working with the architect

The HNWI Intelligence Engine is available in three engagement modes. A founder, fund, or boutique advisory firm can commission a custom build modeled on this architecture and tuned to a specific jurisdiction mix, signal taxonomy, and scoring thesis — useful when the standard taxonomy doesn't fit the operator's market. An existing data, compliance, or capital-introduction team can extend their own platform with the registry-scraping, scoring, or graph-mapping layers as integrated modules. And firms running adjacent products can engage in strategic advisory on registry-monitoring methodology, multi-jurisdictional scraping strategy, and the scoring-model design that turns raw filings into prioritized outreach. Reach out via sintegrium.io or LinkedIn for a 30-minute scoping call.

---

Built by Yurii Staryk · Solution Ecosystem Architect