# We build careful software for the money that financial institutions promise their customers.

Devanka is a small software company being incorporated in British Columbia, Canada. Our first product, Ankasum, is being built to help banks and credit unions define a bonus offer once, check it before launch, and pay it exactly as the published terms say. It is in development, and we are looking for design partners.

[Read about Ankasum](https://ankasum.com/) [Talk to us](https://devankahq.com/contact/)

## What we build

A checking bonus looks simple on a web page: open an account, set up direct deposit, get $300. Behind it sit eligibility rules, date windows, household limits, a budget, a clawback clause and a disclosure that must say the same thing as the code that pays. Today that logic often lives in tickets, hand-written queries and spreadsheets.

Ankasum keeps it in one written definition. That definition drives the plain-English readback, the tests, the approval record, the payout file and the evidence.

[How Ankasum works](https://ankasum.com/how-it-works/)

## How we work

How Ankasum is being built:

Evidence before claims  
We say a control exists when we can show it. Our security and trust page separates what is in place from what is planned, and today almost everything is planned.

Your data stays with you  
Customer and member data will sit in an isolated environment for each institution. Our shared service will hold offer definitions, settings and totals over at least 10 customers. No identifiable customer data will be sent to any AI model.

Code computes, people decide  
Language models may draft text. Deterministic code will validate and compute every amount. A named person at the institution will approve every offer and every adverse decision.

Configuration, not custom code  
A new institution, file feed or data layout will be set up through configuration and versioned definitions, not a fork of the product.

Bad data stops, it does not slip through  
A bad record will be quarantined. A bad file will hold only its own feed, and payouts that depend on it will wait.

## Where the data lives

Approved offer definitions go to the institution's data plane. Only run metadata and totals over at least 10 customers come back to the Devanka control plane.

Institution's data plane, one per institution, isolated, holds:

- daily file feeds
- customer and account records
- rule engine and payout checks
- fulfilment files
- evidence

Devanka control plane, shared software with separated data, holds:

- offer definitions and disclosures
- configuration, users, approvals
- run metadata
- totals over at least 10 customers

Approved offer definitions go from the control plane to the data plane. Run metadata and totals over at least 10 customers go from the data plane to the control plane. Record-level data never crosses this line.

## Where we are

In development

Ankasum is being built. It is not in production at any institution, and we have no customers yet. We are starting conversations with banks and credit unions to test our assumptions, and we are looking for one design partner to run a pilot with us.

[Design partner program](https://ankasum.com/design-partners/)

---

Source: https://devankahq.com/  
Updated: 2026-11-12
