Predvix Limited is a Hong Kong technology company developing Predvix Alpha OS, an autonomous quantitative research operating system. Predvix Labs houses smaller software experiments outside the flagship.
The company's core product: an AI-native research system for algorithmic trading, in development with design partners. The product page covers the research loop, the validation discipline and who it is built for.
Smaller software experiments with a home of their own, so they can ship and learn without being presented as strategic pillars.
AI can now generate code, summarize research and run experiments at a speed that was impossible for small teams only a few years ago. That creates a new problem: research output can scale faster than a human can verify it. Predvix is building Alpha OS around both sides of that shift—greater research capacity and stronger research discipline.
Search across research, market data, events and prior experiments for questions worth testing.
Translate hypotheses into controlled, auditable backtests rather than opaque AI output.
Actively search for leakage, fragility, unrealistic assumptions and failed replications.
Preserve evidence and failures so future work can build on what the company has learned.
Predvix was founded by KK Kwok, a former investment-bank trader, after seeing first-hand how much institutional research depends on better information, disciplined tooling and repeatable processes. Alpha OS applies that perspective to the new problem created by autonomous AI research.
Predvix operates as an AI-native one-person company today. Specialized AI agents and reusable software infrastructure increase development and research capacity, while the product roadmap remains deliberately scoped around high-leverage technical primitives rather than recreating a conventional large organization prematurely.
A polished result is not enough. Important conclusions should remain connected to assumptions, experiments and limitations.
Models and agents are execution infrastructure. Product value should survive individual model vendors and framework trends.
Research objects, experiment history, evaluation data and failure memory should make the system more useful over time.
Alpha OS is in development. We prefer design partners, real experiments and measured progress over inflated launch claims.
We welcome conversations with quantitative researchers, design partners, strategic partners and the Hong Kong technology ecosystem.