The risk intelligence layer between research and capital allocation.
Deterministic portfolio diagnostics and strategy due diligence for systematic funds, multi-manager platforms, and institutional allocators. Math decides. AI explains.
Deterministic scoring across Sharpe, drawdown, robustness battery, fragility index, and walk-forward validation. No black boxes — every number is reproducible.
02 — Monitoring
Is this strategy degrading?
Rolling Sharpe drift, drawdown envelope violations, bootstrap instability, and confidence collapse detection. Know before your risk manager does.
03 — Portfolio Intelligence
How does it interact with our portfolio?
Correlation graphs, strategy clustering, redundancy detection, and diversification scoring. Understand what you're actually allocating to.
What's inside
Every tool an allocator needs. Nothing they don't.
⚡
Deployability Score
A single deterministic score from 0–100 that penalizes flag severity, sample size, history length, and confidence collapse. Reproducible across runs.
🔬
Robustness Battery
Bootstrap resampling, rolling window analysis, outlier stress tests, and regime sensitivity checks. Know if performance holds under pressure.
📊
Walk-Forward Validation
Train/test rolling windows with OOS Sharpe, consistency score, and degradation signals. The closest thing to live track record analysis on historical data.
🧬
Fragility Index
Composite fragility score measuring backtest stability under deterministic stress signals. Identifies strategies that look good on paper but fall apart in deployment.
🗂
Portfolio Clustering
Correlation-based clustering detects redundant sleeves. Know when four strategies are really two. Overlap risk scored and flagged automatically.
🤖
Quant Copilot
Ask natural language questions about your portfolio. Powered by Gemini with full portfolio context. "Which strategy should I cut first?" gets a real answer.
Math decides. AI explains.
Every metric is computed deterministically — the same inputs always produce the same outputs. No stochastic scoring, no model-dependent verdicts. AI commentary interprets the numbers but never generates them. Every report includes a dataset hash and deterministic provenance signature for full auditability.