About this Research OS
This is a dynamic research knowledge base for organising, exploring and presenting a Data Science research process end to end — not a portfolio in the usual sense, and not a general analytics dashboard.
Its job is to make one path visible and traceable: discovery, data, problem, research question, experiment, evidence, finding, impact.
“I found a real problem, found evidence that could answer it, designed valid research, tested the hypothesis reproducibly, then turned the result into something usable.”
That's the difference this Research OS is built to show: research capability and systematic inquiry, not just model-fitting.
Paradigma Riset Dua Lensa (Dual-Lens Framework)
10 Core Principles
- Research is the primary object, not the model.
- Problem ↔ data co-discovery is the main discovery principle.
- Every important claim must trace back to evidence or a source.
- AI helps with discovery and execution, but research decisions stay auditable.
- Popularity or interest does not automatically mean research importance.
- Correlation is never shown as causation without a valid causal design.
- Every experiment must be reproducible.
- Data provenance stays visible.
- Public presentation stays understandable without losing technical rigour.
- Content and presentation are separated, so research can change without a redesign.
Research lifecycle
Promotion from candidate to project is always an explicit decision, never automatic.
Built for an AI research assistant too
The assistant does not edit the interface directly. It produces structured records — a topic, a signal, a candidate, a dataset, a research question, an experiment, a finding — and the web stays a presentation and exploration layer over the same knowledge base a human researcher edits.
{
"type": "finding",
"project_id": "food-price-volatility",
"statement": "Food-price volatility differs substantially across regions...",
"confidence": "high",
"evidence_ids": ["ev-01", "ev-02"]
}This keeps the project auditable, modular, and portable across LLMs and frontend renderers.
What this isn't (yet)
This early version is not a generic CMS, a BI dashboard, a notebook platform, or a collaborative multi-user tool. There's no unnecessary database bloat — content is structured and reproducible, and the frontend never hardcodes research content.