Synthetic Societies.
Informed Decisions.
Explore audience reactions and review campaign drafts before you publish, with synthetic research grounded in Hong Kong demographic data.
Create an accountMarket Research is Biased by Default.
Inherent Human Bias: Self-reported data is fundamentally flawed. Social desirability and recall bias mean consumers rarely do what they say they will do in a survey.
Statistically Disconnected: Traditional focus groups sample tiny, unrepresentative pools that fail to reflect complex, fast-moving regional demographics.
The Impact: Brands make multi-million-dollar strategic bets on aspirational feedback rather than objective economic reality.
Product 01 //
Synthetic Societies.
From public data to enterprise precision — our engine bridges raw public demographics with your private customer intelligence to create an unbiased predictive sandbox.
Phase 1 — Ground-Truth Census Ingestion
We parse raw HK C&SD Census data. Our autonomous AI agents are mathematically structured to mirror exact real-world demographic distributions (Age, Income, Housing, District).
Phase 2 — Private Enterprise Synthesis
Supercharge simulation accuracy by feeding your CMS/CRM data — actual consumption records, payment histories and behavioral preferences — securely into your agents' cognitive loops.
Phase 3 — Behavioral Calibration
Calibrating base census math with your first-party customer records eliminates generic LLM drift. The result is a hyper-targeted synthetic population that thinks, spends and reacts like your true customer base.
Product 02 //
Pre-Flight Check.
Operational brand insurance for high-volatility markets. Before you click Publish, pass your draft copy, campaign assets or press releases through the Pre-Flight Engine to predict sentiment, surface hidden PR-crisis risks and forecast engagement.
The Macro-Baseline — HK Crisis Database
Our agents are continuously trained on a historical dataset of Hong Kong's marketing and corporate PR crises — an algorithmic social immune system that flags linguistic triggers, cultural missteps and high-volatility topics that have historically sparked public backlash.
The Micro-Calibration — Your Audience DNA
Point the engine at your brand's social pages (Facebook, Instagram and more). By ingesting your page's actual content and audience interaction patterns, the simulation adjusts its prediction to how your specific followers will respond — not just the general public.
The Result
Objective, mathematically simulated audience perspectives — a full reaction feed, PR risk scores and ad-delivery forecasts — that cure internal creative blindness and stop PR disasters before they hit the press.
From Public Data to Predictive Power.
Our four-step pipeline ingests raw reality and outputs actionable foresight.
1. Ingest
Parse raw HK C&SD Census data (Age, Income, Housing, District).
2. Spawn
Sample from a bank of ~180,000 persistent, census-mirrored AI personas via our async orchestrator.
3. Simulate
Agents use Chain-of-Thought budgeting to react to your brand's scenarios.
4. Analyze
Output statistically validated predictions on population market behavior.
Technical Architecture.
A transparent look into the Vitapia engine. How we bridge raw demographic data with high-concurrency LLM cognition.
Temporal Knowledge Graphs (GraphDB)
Competitors simulate agents in vacuums. Vitapia agents live in a massively interconnected Temporal Knowledge Graph (Neo4j-backed). Every agent is a node whose edges — neighbours, coworkers, friends — are generated from real demographic overlap (district, age, income) and weighted by a 1–100 influence score, so high-influence agents ripple harder through their network.
This allows us to simulate Network Contagion. If an agent 'buys' a product — or a PR crisis erupts — that state change ripples through their local graph network, dynamically altering the behavior of their synthetic peers to mathematically model viral marketing, social proof and viral backlash.
Neuro-Symbolic Agent Architecture
Relying purely on LLMs for financial predictions introduces massive hallucination risks. Vitapia bridges deep learning with hard algorithmic logic via a Neuro-Symbolic Engine.
The "Symbolic" layer is a deterministic engine that strictly computes each agent's disposable income — progressive salary tax and district-level rent adjusted for housing type — down to the exact dollar from HK C&SD economic data. The "Neuro" layer (the LLM) is then only permitted to calculate the emotional and psychographic variance on that remaining disposable income; a hard post-computation interceptor rejects and retries any economically impossible decision.
Continuous SLM Fine-Tuning (DPO)
Generic API wrappers are easily cloned. Vitapia's ultimate moat is our proprietary, ever-growing training data.
Every economically impossible decision our interceptor rejects — and every deviation our Empirical Validation Engine measures against client-supplied ground truth — is captured as a preference pair in a growing Direct Preference Optimization (DPO) dataset: the training fuel for our localized Small Language Models (SLMs). Over time, the Vitapia "HK-Brain" becomes an impenetrable, highly specialized economic and sociological mirror that generic commercial APIs cannot match.
Proven Accuracy.
We simulated 100 synthetic HK residents against the real 2019/20 HK Household Expenditure Survey. Without manual prompting, our AI agents mathematically matched real Hong Kong human spending behavior.
Scroll to Zoom • Drag to Rotate • Hover for Details
Calculated using Mean Absolute Error (MAE) across 4 real-world data points vs 100 synthetic agents.
Formula:
100% - (MAE = 0.028) = 97.2%
In Food Expenditure Share prediction vs. Real HK Ground Truth.
Evaluated across 4 non-food spending categories (Transport, Services, Goods, Clothing) using MAE.
Formula:
100% - (MAE = 0.108) = 89.1%
Successfully predicting the dominance of miscellaneous services in HK budgets.
Deep Demographic
Roleplay.
Beyond Basic RAG: Injecting raw data makes AI robotic. We give agents Household Size and Occupation to trigger latent socio-economic reasoning.
Asia-First Focus: Tuned specifically for Greater Bay Area cultural and linguistic nuances (Cantonese/English mix).
Competitive Edge: No direct commercial rivals in Asia; academic models lack enterprise speed.
From Question to Insight, in Hours.
Marketing & Growth Teams
Pressure-test campaigns, launches and pricing on a synthetic population before spending real budget — Pulse Check reports land within 24 hours.
PR & Communications
Run every high-stakes post or press release through Pre-Flight Check to catch cultural missteps and crisis triggers before they go live.
Founders & Strategy Owners
Simulate demand, price sensitivity and what-if scenarios without recruiting a single panel — iterate as fast as you can ask questions.
B2B SaaS for
Decision Dominance.
Tier 1: Pulse Check
Pay-per-simulation operational insights.
- ✔ Synthetic Societies: test a launch or pricing change on agents sampled from a ~180,000-strong census-mirrored population
- ✔ Pre-Flight Check: validate a high-stakes post against the HK Crisis Database
- ✔ 24-hour verified insights
Tier 2: Enterprise Custom
Monthly SaaS Subscription.
- ✔ Private sandboxes: upload CMS/CRM records & link your social channels
- ✔ Dynamic simulations optimized with your exact customer DNA
- ✔ Scenario testing API that plugs into your marketing & PR workflows
Infinite Scale.
Limitless Forecasting.
We are architecting the future of urban simulation. Moving beyond basic consumer survey analytics into massive, multi-variable macro-level city modeling.
Cloud Optimisation
Our simulation core runs on elastic, auto-scaling cloud infrastructure. Already provisioned with a ~180,000-persona census-mirrored bank, our asynchronous orchestrator scales horizontally to simulate hundreds of thousands of concurrent AI agents, effectively mirroring entire city demographics in real-time without computational bottlenecking.
Macro Forecasting
Survey prediction is just our baseline. By equipping our synthetic populations with physics engines and historical macroeconomic data logic, Vitapia is expanding vertically. We will offer City Development Analysis (predicting traffic and commercial shifts due to new MTR lines) and Economic Forecasting (modeling district-level retail inflation resilience).
Questions, Answered.
How marketers and business owners use Vitapia to test faster and publish safer.
You describe a scenario — a product launch, a price change, a policy poll — and we run it against synthetic Hong Kong residents sampled from our ~180,000-strong census-mirrored persona bank. Each agent reasons over its own real computed budget and its social circle before deciding. You get a full report with demographic breakdowns and an executive summary within 24 hours, versus the weeks a survey or focus group takes.
Ethical AI for Hong Kong's Future
PDPO Compliant: Synthetic data means zero personal data privacy risks.
Bias Mitigation: Mathematical validation ensures diverse, representative populations.
Ecosystem Contribution: Supports HK's "AI for All" strategy by creating local jobs in agentic AI.
Ready for
Decision Dominance?
Create an account and verify your email to receive welcome credits. Start with a Pre-Flight Check from your dashboard.
Create an accountAlready have an account? Log in