A two-layer architecture built for numbers you can trust.
The Superlytix Engine executes every step deterministically on your full dataset. Above it, the Proposal Layer, Specialized Super Intelligence, suggests what to build from column signatures alone. Nothing it proposes reaches your report until the Semantic Compiler has verified it against your data.
Two layers. One boundary.
The engine works on your rows; the Proposal Layer only ever sees what crosses the RowGuard boundary: column signatures and computed findings going up, a candidate plan and a narrative coming down. You approve every step along the bottom.
Five passes over every row
Deterministic, repeatable and logged: the same file always gives the same model, and every pass is replayed when your data refreshes.
Signal Profiling
Reads every column of every row: types, ranges, blanks and distinct values, in English and Vietnamese number and date formats.
Sanitation Pass
Fixes the structure (title rows, two-row headers, merged cells, totals, placeholders) and the formats, and logs every change it makes.
Topology Inference
Finds the facts, the lookup lists and the keys between them, and builds a star schema with its own Calendar.
Semantic Compiler
Checks every proposal against your data. One metric definition becomes the DAX, the preview number and the plain-language explanation.
Artifact Synthesis
Writes a validated Power BI project, an Excel workbook and a PDF business review with a risk scan.
What we never compromise
Propose–Verify Loop
Every metric, breakdown and chart the Proposal Layer suggests is checked by the engine: do the columns exist, are the types right, does it calculate on your data? Anything that fails is fixed or left out.
Single-Source Semantics
Each metric is defined once in our own formula language. The Semantic Compiler turns that one definition into the Power BI DAX, the preview numbers and the plain-language explanation, so all three always agree.
RowGuard Isolation
The Proposal Layer sees column names, types and short summaries, never your rows, and nothing is used for training. On the Free plan nothing leaves the engine at all.
What we're building next
Live Source Connectors and Drift Detection come with Pro and Enterprise as they ship.
- IN DEVELOPMENTLive Source ConnectorsBuild straight from SQL Server, PostgreSQL, MySQL and BigQuery instead of exported files.
- IN DEVELOPMENTDrift DetectionLoad next month's data and see which numbers moved and which columns changed shape.
- IN DESIGNRule SynthesisThe Proposal Layer drafts cleaning rules (standard spellings, formats, encodings) that the engine tests on every row before you approve them.
- IN RESEARCHMetric Genome LibraryA curated library of tested metrics and report blueprints that grounds the plan in what works.
See it on your own data
Upload a file, and the engine shows you everything it did before anything is built.