Generating risk results is only half the battle. Cortex makes the answer accessible to everyone on your desk — and every result traces back to exactly how it was computed.
Every AI tool that made risk easier did it by making the numbers harder to defend. Cortex refuses the trade. Accessibility goes up. Reliability does not go down — because the model explains the analysis, it never produces it.
Cortex is different by construction — not by promise. The heavy computation stays on your deterministic engine, so cost is bounded, results reproduce exactly, and a named person approves every plan before it runs.
Ask the risk team why a number moved, and you pay in delay, opacity, and disruption — so you ask less than you should.
Ask in plain language or by voice. Cortex finds the path through the system in seconds — and shows you exactly how it got there.
"We have a sophisticated risk system, but I rarely use it directly — reaching an answer means asking the tech team and waiting. I just want to ask a question and get a reliable answer."
— a risk officer at a large asset manager
The three capabilities Cortex leads with — one governed system, grounded in your own configuration and computed results.
A portfolio manager describes a market stress they're worried about. Cortex turns it into a governed factor plan — and explains the result down to the security.
"I'm worried about a sharp equity sell-off with rates backing up and credit spreads widening."
Plain language, the way you'd actually say it — no factor codes, no RiskON expertise required.
It maps your narrative onto the factor ladders your engine actually supports — equity, rates, credit, commodity, volatility — and selects specific, available shifts: EQ −40%, IR +300bp, CS +300bp. It flags honest gaps too, like when no clean historical anchor exists for the regime you described.
You see exactly which shifts and historical anchors it chose, and why. Run it as-is, dial the severity up or down, add a factor, or save it as a reusable scenario — you stay in control of the stress that gets applied.
Results come back immediately. Then drill into any position to see precisely how it repriced under the shocks — a security-level Explain that turns a P&L number into a defensible answer, not a black box.
From a sentence to a governed result — and it shows you how it got there.
"Stagflation: equities fall, rates and commodities surge." Cortex turns a sentence into a deterministic stress plan — the right risk factors, shift sizes, and historical anchors — and shows its reasoning before anything runs.
Say it or type it: "Run a tail-risk decomposition and show me where the loss concentrates." The agent navigates Cortex, runs the analysis, and narrates each step — so you never have to know which module, configuration, or query to reach for.
A portfolio manager thinks in theses, not mapping rules. You shouldn't have to translate your view for an analyst — or wait while someone else does.
The translation is on screen, not outsourced. Approve it, correct it, and move on — in the time it takes to ask.
The people who defend, configure, and run the risk process every day — each with a different challenge, the same answer.
A $20B multi-strategy firm had used a leading commercial risk system for years, but the pace of new strategies and security types outgrew the original implementation — sophisticated instruments like OTC derivative hedges weren't being modeled, and simple proxy rules weren't enough. Within a year, mapping rules were implemented across all security types and the upstream gaps were closed. Today the firm has a firmer grasp of portfolio risk, agile control over model configuration, and full transparency into the process.
The deterministic engine produces the numbers. Cortex explains them — and no single model's mistake reaches you unchecked.
A fast layer sends each question to the right documents and data — no keyword guessing.
A reasoning layer reads the full source material and computed results to produce the answer.
A separate, independent layer verifies every claim against the evidence before it reaches you.
Most firms want AI value but stall on cost uncertainty and internal resistance. Cortex is built to remove both.
Cortex is deterministic-first. Analytics, lineage, and stress run with no LLM call at all — a model is invoked only when reasoning or explanation is actually needed. You're never billed for tokens you don't use, and cost tracks value, not clicks.
The engine is deterministic and unchanged — AI explains results, it never computes them. Every answer is cited to the evidence and independently validated. Nothing reaches the desk on faith.
Cortex frees analysts from being the human runbook and lets your experts scale — it codifies their judgment, it doesn't overrule it. The people who were the bottleneck become the authors.
Runs inside your security boundary — on-prem or private cloud. You decide what the AI can see. Model-agnostic, so you're never locked to a single provider.
Risk environments never hold still — drift is inevitable. The only choice is whether it's invisible or governed. Cortex makes each change a testable event, not a fragile workaround or a long investigation.
New model versions, new instrument coverage, vendor updates, and evolving internal conventions.
VaR horizon and confidence, stress definitions, volatility and correlation methods, proxies, and overrides.
New hierarchies, sleeves, strategies, benchmarks, legal entities, aggregation views, and reporting cuts.
Role churn and ownership shifts — without losing the institutional knowledge that explains the book.
Your desk's needs change. Extend Cortex by asking — new functionality, a fix, a new analytic — and every change is governed, regression-tested, and never touches the deterministic core.
Describe a change in plain language; an agent builds it, runs the full regression suite, and it ships only when it passes. You see it work in a version you test first — nothing reaches production on a promise.
Request a change, new functionality, or a fix — by voice or chat. It's built and regression-tested under that same loop, then delivered as a version you test first. Nothing reaches production until you accept it.
The build agent runs under a fixed set of rules and skills that no one can change — not you, not us in the moment, not the agent itself. The deterministic core stays protected, and Red Swan maintains that governance so extending Cortex never adds risk.
We've been separating signal from noise in institutional portfolio risk since before the Global Financial Crisis.
Red Swan has implemented, managed, and supported risk analytics and models for sophisticated hedge funds for 18 years. Our partners previously led research and analytics teams and institutional risk-aggregation businesses. Cortex embodies that practitioner experience and expertise — and persists your institutional risk intelligence so it doesn't walk out the door when people do.
Experience implementing institutional risk — codified into the system, not dependent on staying in the room.
Implemented, managed, and supported risk analytics and models at Red Swan for 18 years. Prior experience supporting roughly 200 hedge-fund clients on an institutional risk platform.
LinkedInLead risk developer for risk-integration implementations and modeling at Red Swan for 15 years. Portfolio and risk management since 1988.
LinkedInPartner and CTO at Red Swan for over 7 years. Former Head of Research, Analytics & Data at a leading institutional risk firm for over a decade, and lead developer of the analytics and reporting layer.
LinkedInScope, governance, integration, and deployment — the boundaries that make this durable in production.
If your environment has drift, complexity, or key-person dependency, the next step is a scoped discussion: what's changing, what breaks, and what "governed" must mean in your operating model.
