Built for the teams where Aladdin is out of reach and ChatGPT is banned by compliance. Alpha Quant Agent delivers governed workflows with deterministic execution, human approval checkpoints, and audit-ready output.
Live platform — no implementation project required
The analytics catalog is in production. After a workflow diagnostic and onboarding session, your team runs its first governed workflow immediately — no multi-month integration timeline.
Standard reports produced every cycle, not rebuilt manually
Monthly factsheets, quarterly LP letters, GIPS composites, and risk committee packs — generated from a governed workflow, not assembled by hand.
Your proprietary models, governed and ready to act on
Register your house models — MATLAB, R, VBA, or Python — directly into the platform. They run with the same approval gates, anomaly detection, and audit trail as every built-in method. Your IP stays in your Azure tenant.
architecture
Public overview of the workflow. Detailed runtime behavior remains available after login.
01
Bring in data and methods
Load the portfolio inputs and preparation baseline needed for the workflow.
02
Plan or approve the workflow
Choose fast execution or review-first planning based on stakeholder and governance needs.
03
Review outputs and artifacts
Receive analysis artifacts, method notes, and report-ready deliverables for stakeholders.
Direct execution
Move quickly when the request is clear and the workflow can proceed in one governed flow.
Plan-first execution
Review the proposed plan first, then approve execution when additional control is required.
Every method is deterministic, version-controlled, and runs through the same approval gate.
Annualised risk-adjusted return ratios
Compound growth and cumulative return chart
Depth, duration, and recovery per episode
Regime classification: low / medium / high vol
Tail-risk at configurable confidence level
Investor journey and downside deviation metrics
Dynamic volatility model for risk overlays
Bootstrap + regime-switching simulated paths
Your team's proprietary models are often the most trusted signals you have. The problem isn't the model — it's that the output sits outside any governed process. Once registered, your model runs deterministically, its output is compared against your own run history, and every execution is permanently recorded.
Client-registered models run in an isolated subprocess with a strict import whitelist. No network access, no filesystem writes.
After every run, three intelligence layers activate automatically — no dashboard to open, no threshold to remember to check.
Every metric is benchmarked against your own run history. Anything beyond 2σ is flagged with context: '3.3σ below your 14-month average.' Activates after 5 runs.
Set beta caps, drawdown floors, VaR ceilings once in Settings. Every run checks automatically. Breaches are flagged, recorded in the audit trail, and surfaced in the Flags tab.
A plain-English commentary is generated from the computed numbers only — not from general knowledge. Every sentence cites the metric it came from. Defensible to a compliance officer.
Every governed run produces a structured report. Here is what a fund teaser pack output includes — the full report is shown on request.
Two sample reports from this workflow are available to download below — generated from anonymised fund data.
See it on your dataGenerated from anonymised fund data. Open in browser or download — no signup required.
Animated preview
Inside the report pack
Auto-playing pages from the same PDFs linked below.
Hover to pause | Click a dot to jump
Fund Teaser
1 page · A4 landscape
Single-page snapshot with 5 core KPIs, risk-adjusted verdict, and trailing returns. Sent to prospective LPs before the first meeting.
Investment Pitchbook
9 pages · A4 portrait
11-year track record with peer universe comparison and investor suitability profile. Used in LP due diligence and initial allocation meetings.
Monthly Factsheet
8 pages · A4 portrait
Full performance attribution, rolling risk metrics, and fund-vs-benchmark analysis. Standard monthly reporting pack for institutional investors.
Investor Letter
7 pages · A4 portrait
Narrative-driven quarterly letter with market commentary, attribution analysis, and forward outlook. Designed for sophisticated LP communication.
Risk Oversight Report
8 pages · A4 portrait
COVID crash and rate shock stress scenarios with mandate compliance table. Used in monthly risk committee and board reporting.
Before any sensitive analysis runs, Alpha Quant Agent pauses and requires an explicit human decision. This is not a UI suggestion — it is enforced at the API and worker level. Every approval is recorded with a timestamp and decision log.
Plan proposed
The agent proposes a structured JSON plan listing every method, parameter, and execution order.
Human approves or rejects
Execution cannot start without an explicit approval decision. Rejections are logged and recovery is prompted.
Execution + full audit trail
Methods run deterministically. Every step, decision, artifact, and timestamp is written to the audit record.
Regulatory bundles are force-gated to planned mode. The worker enforces this — no client-side override is possible.
Azure Front Door + WAF
DDoS protection · OWASP rule sets · TLS 1.2+
All traffic filtered before reaching the platform
CSV, Excel, or API feed
Azure Blob Storage
Isolated worker run
PDF report + raw data
Data region determined by your location
No data leaves your jurisdiction
Entra ID
Identity & access
Blob Storage
Client-isolated files
Container Apps
Isolated execution
PostgreSQL
Run logs & audit
Next step
Begin with a workflow diagnostic, then expand into custom research engineering, workflow automation, or Azure deployment as needed.