Three pains every sales and BD team hits weekly. Each one is what your reps actually complain about, not what a feature page would call them.
Half the strategies that ship with a 2.1 Sharpe in the backtest drop below 0.8 within six weeks of paper trading, and the post-mortem catches the same overfit every quarter.
Execution cost is estimated as a flat 5 bps because nobody on the desk has time to decompose spread, impact, and adverse selection per strategy, so live P&L misses target by 30 to 60 percent.
Risk reporting is a Friday-night Excel scramble: parametric VaR in one workbook, historical VaR in another, stress tests in a third, and the CRO signs off blind.
Every pipeline below is a shape you wire on the canvas using the crew and tools further down. Not a feature we ship for you, a pattern you configure.
Run IC, Sharpe with standard error, alpha-decay half-life, and parameter sensitivity on a candidate factor, applying Bonferroni or Benjamini-Hochberg correction across the tested grid. Static context holds the factor library and the desk's significance thresholds.
Take the backtest fills and decompose cost into bid-ask spread, market impact, and Glosten-Milgrom adverse selection. Scoped database tool reads tick captures only, no broker writes possible from this workflow.
Compute fractional Kelly at 0.25x to 0.5x against the live book using a Ledoit-Wolf shrunk correlation matrix. HITL gate blocks the allocation change until the PM approves the size, the correlation set, and the volatility target.
Run parametric VaR, historical VaR, CVaR, and Cornish-Fisher tail corrections, then replay the 2008, 2020, and 2022 stress books. Cross-run memory carries circuit-breaker thresholds from the prior review so the limits do not reset each Monday.
Score Deflated Sharpe Ratio against the trials count, demand walk-forward validation, and refuse any single-regime backtest. The rag_retrieve tool pulls prior failed strategies so the same overfit pattern does not pass twice.
Synthesize the five specialist sections into a memo with an ACTION line of DEPLOY, PAPER_TRADE, BACKTEST_MORE, PAUSE, or KILL and exact recommended params. Replay captures every tool call so compliance can audit the verdict end to end.
Real personas from the quant_firm crew. Each ships with a tuned system prompt and a default tool allowlist. Swap models per persona on the canvas.
Runs signal-quality analysis on a strategy: IC, Sharpe with standard errors, parameter sensitivity, and Bonferroni or Benjamini-Hochberg correction across the tested combinations.
Decomposes execution cost into bid-ask spread, market impact, and adverse selection, then quantifies the conditions under which slippage and latency wipe out the edge.
Sizes the strategy with fractional Kelly and a Ledoit-Wolf shrunk correlation matrix, then decides whether it earns a slot next to the live book.
Computes parametric and historical VaR, CVaR, and Cornish-Fisher tail corrections, runs the 2008, 2020, and 2022 stress books, and sets hard circuit breakers.
Scores overfitting with the Deflated Sharpe Ratio, refuses any strategy without walk-forward validation, and issues an APPROVED, CONDITIONAL, or REJECTED verdict with the exact failure conditions.
Distills the five specialist sections into one PM-ready memo with an ACTION line of DEPLOY, PAPER_TRADE, BACKTEST_MORE, PAUSE, or KILL, and exact recommended parameters.
Every tool below is a real shared tool from the Melaya bundle. Allowlist per agent; HITL-gate the writes; revoke any of them in one click.
Pull price history, technical indicators, FX rates, and macro series for factor research and out-of-sample regime tests. Read-only by design, no broker writes possible from this bundle.
Build the per-workflow knowledge store from strategy specs, prior tearsheets, post-mortems, and risk policy docs. Powers the three knowledge layers QuantAnalyst and QuantSkeptic call against. Writes to the workflow store only, never to the live trading system.
Read backtest output, tick captures, and PnL attribution from your warehouse. The destructive sql_execute and sqlite_execute calls are HITL-gated and off by default for this crew.
Extract tables from research PDFs, vendor tearsheets, and prior risk filings into structured inputs the QuantAnalyst can cite. No writes; reads only the files you point it at.
Validate backtest CSV exports, summarize distributions before they hit the analyst, and hash result files so the audit log can prove the tearsheet under review is the same one the PM signed.
Read PM-facing Excel risk books and emit the final deploy-or-kill memo as a Word doc the compliance archive can ingest. All writes are HITL-gated and stage to a draft folder, never overwriting the prior version.
Drop the ReportWriter's deploy-or-kill memo into the desk channel once a human has approved it. HITL by default, with the same audit log as every other write.
Every pipeline ships with three layers of knowledge access. Mix and match per agent on the canvas. No shared vector space with another tenant, no surprise reads, no opaque retrieval.
includeContextPer-pipeline documents appended to specific agents' input on every run. The ICP brief, playbook, pricing sheet, or won-deal email corpus. Whatever needs to be there before the agent thinks. You pick which personas get which docs.
rag_retrieveA scoped tool granted per-agent. When the agent decides it needs more depth, it queries the workflow's vector store on demand. Same knowledge base as Static context, accessed only when the model asks for it.
pipeline_memoryPipeline-level state that carries from one run to the next. Yesterday's research is in scope for today's follow-up. The crew remembers what it already prospected, what got approved, what was sent. The audit log is the second-order knowledge base.
Sandbox tier is free with no card. Join the waitlist and we will email you the moment a slot opens.
← Back to every use case