In continuous production since January 2025. DMCC member. Dubai-based. Built by an operator for desks running leverage exposure.
RiskLoom is built by an operator with 12+ years across data engineering, platform architecture, and analytics, with an early career on the trading floor of a London proprietary firm. The problem it solves is one the founder lived: by the time a liquidation cascade is visible in price, it is already too late to act on it.
The premise is narrow and deliberate — instrument the structural pressure that precedes forced selling, with enough lead time for a desk to act. Built and run in continuous production since January 2025, the approach has been validated against 314+ documented cascades and counting, every claim reconcilable to a source-hashed, per-minute event in the archive.
RiskLoom is now built for any desk carrying leverage exposure — risk officers, prop traders, market makers, exchanges, funds, and family offices — who need the same warning.
The full methodology is disclosed at evaluation, under NDA. What follows is the high-level architecture — four instrumentations, each reproducible and reconcilable to source data.
Every asset is calibrated against its own 90-day distribution — not one universal band. What's extreme for a large-cap is ordinary noise for a smaller name. This is most of the work, and the difference between a signal a desk trusts and one it mutes.
Each cascade is matched against Hyperliquid within a tight window to confirm it is structural, not a single-venue artifact. Cross-venue confirmation is first-class, per symbol.
Cross-asset clusters validated against a shuffled-onset null model over 10,000 permutations — 4 validated clusters, ~600× over chance, p<0.0001. Measured, not asserted.
Liquidation figures are real, atomic, per-minute forced-liquidation value — not rolling aggregates that flatter the headline. The $1.49B across the 2026 cohort is the smaller, defensible number, chosen on purpose.
The CRI engine goes live, detecting and documenting its first liquidation cascade under real market conditions.
The verified archive crosses 100 reconcilable cascade events, each with source hash and per-minute inputs.
Hyperliquid integration adds CEX/DEX cross-venue validation as a first-class signal.
Permutation-tested cluster detection confirms correlated cascade onset across markets — 4 validated clusters at p<0.0001.
The 2026 cohort is fully reconciled to per-minute liquidation data — $1.49B real notional, 134-minute median advance lead.
Directional sizing layer planned, paired with regime and explicit confidence semantics.
Dubai-based, engaging institutional counterparts across the global financial calendar.
RiskLoom is built by a team with hands-on backgrounds in data engineering, quantitative research, and institutional crypto trading. Based in Dubai, DMCC member, operating through 2026 and beyond. The product is held to one standard: every public claim reconciles to a documented, reproducible event.
Every claim is reconcilable to SQL. Numbers shown anywhere can be traced to the data that produced them.
Source hashes, manifest-backed outputs, reproducible during evaluation. Built for diligence.
Detection and lead time only. RiskLoom does not forecast price or generate trades.
Same inputs always produce the same outputs. No hidden state, no silent drift.