What it does well
It fits your systems and models exactly, and your team owns every line, with no vendor to depend on. Some of the largest banks have built their own, such as Citi's XiNG platform.
Firms already keep figures such as exposure against limits current in other ways. Calseda gives each figure its own calculation, rerun only when a change affects it, and opening to the values it used. Here is what each alternative does well, and when to pick it instead.
| The alternative | What it does well | What Calseda adds |
|---|---|---|
| Building it yourself | Fits your systems and models exactly, and your team owns all of it. | The parts every figure needs, built in, so your people write only the calculation. |
| Stream-processing frameworks | Every event processed at high volume, with windows and joins. | A calculation per figure that finds its own inputs, and the lineage of every result. |
| Real-time risk and analytics platforms | Large sets of risk measures, aggregated and drilled into by many users. | Your own calculations, or your own engine, kept current on your own infrastructure. |
| Lineage and data-catalogue tools | A map of data across the whole estate, with definitions and owners. | The exact values behind each figure, captured as it is calculated. |
| Batch reports with intraday reruns | One consistent snapshot at a set time, simple to run and to control. | Each figure recalculated when its data changes, and marked stale when its data is late. |
It fits your systems and models exactly, and your team owns every line, with no vendor to depend on. Some of the largest banks have built their own, such as Citi's XiNG platform.
The parts every figure needs come built into Calseda: connections to the data, recalculation when an input changes, retries, failover over Kafka, sign-in and roles, and lineage. Your quants or developers write the calculation, with our support, and it stays your code under your own controls.
Or keep your engine: Calseda calls it when its inputs change, on their latest values, with a time limit per run.
The better choice when a single figure with a few stable inputs is all you need, or when your firm wants to own every part of the platform.
They process every event at high volume, with aggregations over time windows and joins between streams. Many guarantee each event is processed exactly once, which ledgers and payments need, and operations teams know how to run them.
A stream processor runs one pipeline for every record, set when it is deployed. In Calseda each figure has its own calculation, which finds its inputs from the data (a portfolio, then the prices of what it holds) and rewires when that data changes.
Under load, each calculation and each screen gets the newest value, never a queue of old ones.
The better choice when every event must be processed, or the work is aggregation over windows of events. The two also work side by side: what a stream processor writes to Kafka can be one of Calseda's inputs.
They aggregate risk measures across desk, counterparty, currency and scenario, and let many users drill in. Many come with standard measures and reports ready to use.
Your own code, or the engine you already run, rerun only for what each change affects, in your own environment, on your Kafka.
The same live figures reach browser pages, Excel, Python notebooks and AI agents, each under the person's own sign-in, and a personal what-if runs your production calculation with a deal added.
The better choice when the vendor's standard measures fit, and interactive analysis across many dimensions is the main use.
They cover databases, ETL jobs, files and reports, often down to the column, and show which reports a change to a source will affect before it is made.
A catalogue shows which tables feed which report. Calseda shows the values one figure was made of, as it used them: if a trade has been amended since, the lineage shows the trade as it was, and can show the new version beside it.
A value entered by hand shows who entered it. The map of how data flows is drawn from what the calculations actually read, so it can't drift from the code.
The better choice for lineage across the whole estate, field by field, including the ETL before data reaches Calseda. The two fit together: the catalogue for the estate, Calseda for the figures it calculates.
Every report comes from one consistent snapshot as of a set time, and the controls around it are well understood.
Between runs, figures drift from the book. Calseda recalculates only the figures a change affects, so the cost follows how much changes, not how big the book is. A figure resting on late data is marked stale.
The better choice for figures needed once a day, such as a return as of a reporting date, or figures that must reconcile to the general ledger at a cut-off.
Not a ledger. Calseda keeps the latest value and skips the ones in between. Where every single event must be processed exactly once, as in payments or accounting, keep the systems built for that.
Not a workflow engine. A Calseda calculation follows its inputs; it does not wait for people. Workflows with human approvals belong in a tool built for them.
Run one of your figures on Calseda beside today's process: live in about six weeks, for a fixed fee. At the end you keep the framework, licensed to run what you built on one server.