Software & Data Platforms · Analytics
AI ENGINEERING
Correct SQL from plain English, against a database built to resist it
2-3 days
of analyst queue reduced to seconds.
Software & Data Platforms · Analytics
AI ENGINEERING
Correct SQL from plain English, against a database built to resist it
2-3 days
of analyst queue reduced to seconds.
Industry
Enterprise data and analytics platform, venture backed.
Scale
$1.5B+ valuation (Series D), $215M raised. Users: analysts and business teams querying a large, high-cardinality real-time analytics database.
Engagement
Implementation: a production text-to-SQL system.
Industry
Customer-experience technology company, venture backed.
Scale
$1.5B+ valuation (Series D), $215M raised. Users: analysts and business teams querying a large, high-cardinality real-time analytics database.
Engagement
Implementation: a 6-month embedded build, core build of roughly 12 weeks with 8 to 10 engineers, plus onboarding and standby support.
Industry
Enterprise data and analytics platform, venture backed.
Scale
$1.5B+ valuation (Series D), $215M raised. Users: analysts and business teams querying a large, high-cardinality real-time analytics database.
Engagement
Implementation: a production text-to-SQL system.
Industry
Enterprise data and analytics platform, venture backed.
Scale
$1.5B+ valuation (Series D), $215M raised. Users: analysts and business teams querying a large, high-cardinality real-time analytics database.
Engagement
Implementation: a production text-to-SQL system.
// THE PROBLEM
Where time went
The database was fast because it was unusual: a specialized SQL dialect, approximate-count sketches, unsupported joins. Most non-trivial questions went to a small data team and sat 2 to 3 days in the queue, engineers who should have been building the platform answered one-off queries, and off-the-shelf text-to-SQL returned confident wrong numbers with no error at all.
// THE BUILD
What we built with them
LevelUp built for correctness first, with the data team: schema-aware retrieval so the model cannot invent fields, query planning, dialect-correct generation with the sketch and time functions injected, static validation before execution, and execution-guided self-correction bounded to a few attempts. Approximate results are surfaced as estimates. Underneath it, a correctness benchmark scored on execution accuracy, the number almost no text-to-SQL project measures.
// THE PROBLEM
Where time went
The database was fast because it was unusual: a specialized SQL dialect, approximate-count sketches, unsupported joins. Most non-trivial questions went to a small data team and sat 2 to 3 days in the queue, engineers who should have been building the platform answered one-off queries, and off-the-shelf text-to-SQL returned confident wrong numbers with no error at all.
// THE BUILD
What we built with them
LevelUp built for correctness first, with the data team: schema-aware retrieval so the model cannot invent fields, query planning, dialect-correct generation with the sketch and time functions injected, static validation before execution, and execution-guided self-correction bounded to a few attempts. Approximate results are surfaced as estimates. Underneath it, a correctness benchmark scored on execution accuracy, the number almost no text-to-SQL project measures.
// THE PROBLEM
Where time went
The database was fast because it was unusual: a specialized SQL dialect, approximate-count sketches, unsupported joins. Most non-trivial questions went to a small data team and sat 2 to 3 days in the queue, engineers who should have been building the platform answered one-off queries, and off-the-shelf text-to-SQL returned confident wrong numbers with no error at all.
// THE BUILD
What we built with them
LevelUp built for correctness first, with the data team: schema-aware retrieval so the model cannot invent fields, query planning, dialect-correct generation with the sketch and time functions injected, static validation before execution, and execution-guided self-correction bounded to a few attempts. Approximate results are surfaced as estimates. Underneath it, a correctness benchmark scored on execution accuracy, the number almost no text-to-SQL project measures.
// PROOF POINTS
5 to 7 business days to under 30 minutes
intake to requisition, for a standard catalog buy
5 to 7 business days to under 30 minutes
intake to requisition, for a standard catalog buy
An estimated 15 to 20 percentage points
on-contract spend up, several million dollars a year in recovered negotiated savings on a procurement base this size
An estimated 15 to 20 percentage points
on-contract spend up, several million dollars a year in recovered negotiated savings on a procurement base this size
Roughly 30 to 40%
of buyers' week reclaimed from routing and data entry for sourcing strategy and negotiation (estimated)
Above 90%
of requests correctly commodity-coded at intake, up from an estimated 60 to 70%
2 to 3 days to under 30 seconds
in the data team's queue, self-serve
About 92%
execution accuracy on the correctness benchmark, up from roughly 60% for an off-the-shelf approach
About a dozen to 200+
from SQL-fluent analysts to business users who can ask the database directly
About a dozen to 200+
from SQL-fluent analysts to business users who can ask the database directly
Roughly 40%
of the data team's week returned to platform work (estimated with the team)
Roughly 40%
of the data team's week returned to platform work (estimated with the team)
STACK
Claude
Schema Linking
Pre-execution Validation
Execution-Guided Self-Correction
Correctness Benchmark
In production with the data team; decisions run on current data instead of on whatever the queue delivered days later.
// How we measured
Execution accuracy on a domain gold set of questions with expected results, tracked with latency and cost; time to answer against the workflow-review baseline.
// Build with confidence
Talk to us about testing your agents before launch.
// Build with confidence
Talk to us about testing your agents before launch.
// Build with confidence
Talk to us about testing your agents before launch.
// Build with confidence