Analysis
without
the noise.
A small team in Singapore building tools that help investors read markets more clearly — without the clutter that most platforms add.
Where the idea started
Rimvaro came out of a specific frustration: financial data tools were either too simple to be useful or too complicated to use without a dedicated analyst. Founded in 2019, the project set out to close that gap — not with more features, but with better signal extraction. The focus from day one was on surface-level clarity backed by deep model logic.
What the tool actually does
The platform ingests multi-source market feeds and applies machine learning models trained on structured financial time series. Instead of raw charts, users see interpreted signals — pattern identification, anomaly detection, and context flags. It does not predict outcomes. It reduces ambiguity so that decisions are based on cleaner information rather than noise.
How client data is handled
Each client's data environment is isolated — no shared inference pipelines, no aggregated behavioral profiling across accounts. Analysis runs on dedicated compute per session. Rimvaro operates under Singapore's PDPA framework and applies encryption at rest and in transit as standard practice, not as an optional tier.
The people behind it
A tight group of researchers, engineers, and former analysts who wanted to build something they would actually use themselves.
Tarquin Welbeck
Head of Quantitative Research
Spent eight years building risk models for institutional desks before joining Rimvaro. His focus is on how models behave at market inflection points — the moments where most tools give contradictory outputs.
Signe Olafsrud
Lead ML Engineer
Background in applied statistics and time series forecasting. Signe designs the model architecture and oversees how signals are validated before they reach the interface. She is skeptical of outputs that look too clean.
Dariusz Kwiatek
Product and Client Experience
Works directly with clients to understand how they use the tool in practice — what information they act on, where they hesitate, and what the interface should not do. Most product decisions at Rimvaro start with his notes.
From the floor
What we learned building this
Most early assumptions were wrong. The first version had six signal types; users consistently relied on two. Complexity was cut, precision was improved, and the tool became more useful by doing less. That approach guides every update.
