/ Product

Ask it a question. It reads your ERP export and tells you what changed.

No dashboard to build, no query to write. Type a question the way you'd ask a coworker, and Optima Engine reads your ERP data and tells you what's actually wrong right now.

It runs on your own server, on a schedule your IT team controls. Nothing about your data goes anywhere else — not to us, not to any cloud model.

optima engine
illustrative example
>Is logistics over budget this quarter?

One department is driving the variance. Combined exposure: $365K.

Logistics, FY2024
Highest variance across all departments
-$365KCritical
MAT-01 margin
2% margin — below the 10% threshold
$84K/yrWatch
/ How it works

Three steps. No data team required.

01

Connect your data

Someone on your IT side sets up an export of a handful of ERP tables. It's a one-time setup — after that it's automatic, nobody has to remember to do it again.

02

The pipeline runs continuously

On your own server, in the background. It cleans, joins, and re-aggregates the export into the tables the AI reads from — automatically, on every new export.

03

Ask it anything

Type a question the way you'd say it out loud. Get back a number and what to do about it — not a chart to interpret yourself.

"
The AI doesn't calculate your margin or your budget variance — the pipeline does, in code, before the model ever sees a number. 18 checks run against every batch before a finding is allowed through.
How the numbers stay trustworthy
/ The validation layer

Wrong numbers never reach your leadership team.

Before a batch is written to the layer the model reads from, it runs through 18 checks against the raw data. A few examples:

Cross-validated

Margin matches its own math

margin_pct is recomputed from revenue and cost and compared against the stored value — if they don't agree, the batch is halted before it reaches the model.

Cross-validated

Budget variance matches its own math

budget_variance is checked against actual minus planned for every department. A mismatch stops the pipeline — no silent rounding, no drift.

Structural

No missing join keys

Order, customer, and material IDs are checked for nulls before anything gets joined — a broken join never quietly produces a wrong number downstream.

Your own team can see exactly how each check works — it's not a black box on either side.

/ Also true

The rest of the trust story.

Computed, not guessed

The math happens in code

Margin, budget variance, and delivery metrics are calculated directly from your ERP data by the pipeline — not generated by the AI. The model explains what the pipeline already computed.

Always current

Re-runs on every export

The pipeline reprocesses automatically each time new data lands. No one has to remember to rebuild the report — it's already done by the time you ask.

On-premises only

Never leaves your server

The model runs on the machine it's installed on. There's no step where your data or your findings get sent anywhere else.

See it on your own ERP export.

Access is granted selectively. Every request reaches Akul directly.

Your server No cloud connection No third party