Skip to content
Doing AI

Trust

Check us rather than believe us.

Most of what a stranger uses to judge a supplier — a logo wall, a review score, a certification badge — we do not have, and we are not going to imitate. What we have instead is that every number on this site carries the method that produced it. This page is that method, in one place.

The architecture is the security answer

Your systems run in your accounts.

Most vendors answer a security questionnaire by describing how carefully they guard your data. The stronger answer is not to hold it.

01

Where the system runs

Your accounts, not ours

Everything we build is deployed into your own cloud, on your infrastructure, under your billing. We are not a SaaS vendor holding your operational data in a tenant you cannot see. That is an architectural fact rather than a policy promise: there is no Doing AI datacentre for your records to sit in.

02

Who can reach it

Access you grant, and revoke

Our access is whatever your administrators issue and nothing more, and it ends when you end it. Systems we build are documented so that your own team, or any engineer you hire after us, can operate them without us in the room.

03

If we disappear

You keep running

Two people is a fair thing to worry about, so the answer is written into the agreement rather than left to trust. Source, documentation and data are yours and exportable at any point, and nothing we build needs a licence from us to keep working.

04

What we train on

Nothing of yours

Your data is not used to train external models and is not pooled across clients. Where a system calls a model provider, which provider and on what terms is written into the scope before development starts rather than discovered afterwards.

Every figure, and where it came from

The numbers, with their working shown.

These are taken from monthly engagement reports prepared for each client's own finance and management review — written to be argued with by the people paying for the work, not written afterwards for a website.

Education · admissions and fee operations

The compliance queue that used to be a spreadsheet

A global education group

The contribution figures are reproducible from git history across the twenty repositories, with author identities normalised and bot commits excluded. The report they come from was delivered to the client's own finance and management for review, not written afterwards for a website. We will walk a prospective client through it under NDA.

₹3,14,000

Revenue double-counting corrected

Self-paid orders on the deactivated source system, reversibly cancelled during the migration so finance reconciled against the real records.

15.6%, ranked 1st of 43

Share of all engineering across the platform

Authored non-merge commits across 20 repositories, 1 Jan – 2 Jul 2026, identities normalised and bot commits excluded. Reproducible from source control.

43

Production deployments in the heaviest month

Automated image updates through CI to the production cluster during June 2026.

Education · finance and accounting operations

The ledger balanced until somebody actually checked

A global education group

Each defect is recorded against the commit that fixed it and the production release it shipped in. The client's finance team reviewed and accepted the corrections — the strongest verification available for accounting work, because the party who would object is the party who signed it off.

304 of 304

Recovered payments silently discarded

A payment class matching neither branch of an accepted-types filter, dropped in full — a 100% loss with no error raised and no failing test.

~28,000

Student fee ledgers showing a wrong paid amount

Roughly 95% of all paid ledgers, where the paid column was double-counted platform-wide. Fixed and shipped 31 July 2026.

5

Distinct reporting defects found by reconciliation

Each identified by reconciling the general ledger against source records, not by an alert, an exception or a failing test. None had ever surfaced as an error.

158

Student records migrated, each ledger-verified

Four migration operations, every one dry-run before commit and checked against the live ledger rather than a database read-back. No money moved.

Education · student communications

Adding one WhatsApp message used to mean a deploy

A global education group

The claim that a new campaign needs no release is checkable the plain way: the next institution went live the following month without a single per-event code change, which is recorded in the May report.

3

Institutions live on one shared pipeline

Each with its own template-parameter shape, previously requiring separate hand-written handling.

9 + welcome + document rejection

Student lifecycle events wired

Enquiry, interview cleared and not cleared, fee verified, and the registration and course fee events, each with its own parameter builder.

None

Redeploys needed to launch a new campaign

Defaults merged into configuration, so campaigns are added without touching or shipping the binary. Previously every new event required a code change and a release.

Education · internal operations

Four internal tools became one platform

A global education group

The platform runs in its own namespace with date-tagged image deploys, so its release history is a matter of record rather than recollection. The permission model can be read straight out of the schema.

4 → 1

Internal applications consolidated

A sales intelligence tool, a reporting engine, dashboards and an internal app, replaced by a single platform on its own cluster.

79 permissions across 26 resources

Access control model

A resource-by-action permission system with ten system roles and row-level security, replacing per-application access rules.

Replaced by a workbench

Manual finance reconciliation

Bank-statement to receipt matching moved from spreadsheets and hand-matching to rule-based ingestion with exact-reference matching, delete-with-undo, and suggestions for the remainder. A recurring monthly task, not a one-off.

28

Production deployments in one month

Date-tagged image deploys during July 2026, on a platform sustained across all 31 days of that month.

Sales operations · any sector that runs a phone team

The call data existed. Nobody could act on it.

It is our product, so the whole of it is available to look at — the running system, the API and its history. Nothing here depends on a client confirming anything.

2 → 1 view

Telephony systems unified

Two call platforms previously reconciled by hand, ingested into one schema with sentiment and scorecard reporting on top.

Closed the loop

Calling joined to outcomes

The platform reads the enrolment warehouse directly, so calling activity and what it produced are answerable in one query rather than two exports.

Key-gated public API

Integration surface

Documented read-only endpoints with case-insensitive filtering, so downstream systems consume the data rather than re-entering it.

Reporting and finance operations · any multi-system business

Eleven reports that used to be assembled by hand

Ours to show in full. The report definitions, the warehouse schema and the scheduling are all inspectable, and the outputs are the ones finance actually uses.

7 → 1 warehouse

Sources consolidated

Five operational databases and two payment-gateway APIs, previously exported and reconciled separately for each report.

11

Report types generated automatically

Across four institutions, delivered by scheduled email and self-service download instead of being compiled and sent by hand.

Recurring, not one-off

Reporting effort per period

The saving repeats every reporting period rather than being a single migration — which is the difference between a project and an asset.

Reporting integrity · any business that exports data

The export quietly returned ten rows

A global education group

This is the kind of defect worth judging us on, because it is the kind nobody is paid to find. Ask us on a call what we would look for first in your reporting, and see whether the answer sounds like someone who has been here before.

10

Rows a full export actually returned

The unlimited case fell through to a ten-row default. Reports were built on it and read as complete.

It did not

How it announced itself

No error, no exception, no failing test. The bug was found by someone reading the code path, not by anything the system reported.

Silent truncation

Class of defect closed

Fixed at the default rather than at each call site, so a future caller that forgets a limit gets the answer instead of a fraction of it.

Regulated review · lending, claims, onboarding, compliance

Three pairs of eyes, and a record of whose

A regulated institution

The permission model and the audit schema are the artefact — they can be read directly and either separate duties properly or do not. The institution it was built for is regulated, and accepted it.

3

Review tiers with separate roles

Each tier a distinct role with its own permissions, so a reviewer cannot complete a stage that is not theirs.

Written as it happens

Audit trail

Captured at the moment of each decision rather than reconstructed later, which is the only version that survives being questioned.

Yes

Configurable per institution

Approval rules are configuration, not a code branch — so a second institution with different requirements does not become a second codebase to maintain.

Billing and pricing · any business with discounts or instalments

Nobody could agree what was actually owed

A global education group

The corrections were accepted by the finance team who had been living with the discrepancy, which is the only sign-off that means anything on a billing question.

~11,000

Balances reading zero when money was owed

Applications whose ledger showed nothing payable because a charge type was excluded from the calculation, so the customer could not pay even if they wanted to.

Fixed at the source

Historic orders that stopped resolving

The first price and discount row of each chain was backdated, so orders paid under an earlier price resolve correctly instead of failing.

One

Places the amount owed is now computed

Previously several, each defensible and none authoritative. Disagreement about the number was the defect, not any single wrong answer.

Delivery reliability · any team shipping to production

Nine portals that could not ship themselves

Our own infrastructure, so there is nothing to take on trust — the deployment history shows the gap and shows it closing.

9

Portals unblocked

Each frozen behind the same write-back failure, all released once the underlying permission was corrected.

A token, not the tooling

Root cause

A guest-level access token that could read but not write back. The deployment system, which had absorbed the blame, was working correctly throughout.

Back to zero

Manual steps per release

Hand-bumping had quietly become the routine. The fix removed the step rather than documenting it.

The part most sites leave out

What we do not have.

A supplier who only tells you the good half is telling you half. Here is the other one, and each line stays until it stops being true.

Named clients and logos

None on this site, and not because there are none. Confidentiality is one of the few binding clauses in the agreements we work under, so the name is the client's to give rather than ours to take. Every case study records that the name can be released if that client agrees in writing.

Published testimonials

Ask instead for a reference call. We will put you on a call with someone running what we built, under NDA where their agreement needs it — which tells you more in twenty minutes than a paragraph on a website ever has. We publish a written quote only once a named person has approved the exact wording, so there are none here yet.

SOC 2 or ISO 27001

Neither, today — and the architecture is why it matters less here than it would elsewhere. Systems we build run in your accounts, under your controls, so the questions a certification answers about our infrastructure mostly do not apply to where your data actually sits. We answer security questionnaires directly and in full, and if a certification becomes the condition of a specific engagement we will tell you what it would take and how long.

The engagement reports behind every figure on this site were written for the client's own finance and management review, not for a website. We will walk a prospective client through them under NDA, with identities redacted where the agreement requires it. Ask on the first call.

Who is accountable

Where the people came from.

A client's name is theirs to give. Where we worked before this company existed is ours to state, and it is checkable.

Comcast

XRM analytics data pipelines, through UST and Tulasea

NielsenIQ

Retail and CPG analytics, Chennai — segmentation and sales forecasting

Adidev Industries

Software architect since 2023 — OEM and supplier ERP, Hosur

The commitments these rest on

01Production is the deliverable

02You own what we build

03We build into your existing stack

04The team you meet is the team that builds

05We scope honestly, including when the answer is less