AI
Proof
Pricing
Book a demo
Get the free SAS assessment
Targets
Warehouses
SAS2PY converts DATA step, macros, DI Studio jobs, and SAS/STAT, and emits native PySpark notebooks, Delta tables, and Workflows, with column lineage in Unity Catalog and a parity report your auditors can keep.
Updated
4–6 weeks · about 10K lines of your SAS ·
Deterministic parsers read the SAS estate and emit native Databricks code — not SAS sitting on a cheaper cluster.
SAS programs → SAS2PY parser → PySpark + Delta + Workflows
SAS
Base SAS
DATA step / macros
DI Studio
Jobs + mappings
EG / EM
Projects + flows
Viya / CAS
CASL + actions
SAS2PY Parser
Deterministic parse
AI where it helps
Row-level parity
Before cutover
PySpark emit
Set-based, not loops
Workflow emit
Job DAGs
Databricks
PySpark notebooks
Set-based, not loops
Delta Lake
ACID + time travel
Workflows
Replaces SAS Grid
Unity Catalog
Column lineage
Repos + DABs
CI/CD promote
MLflow
SAS/STAT models
MigryX AI handles the logic parsers cannot resolve alone, and every change it makes goes through the same parity checks. It runs on a model you approve, air-gapped if your estate requires it.
Lakebridge is free and covers SQL dialects. SAS2PY covers the SAS estate.
| Capability | SAS2PY |
|---|---|
| SAS DATA step, macros, %INCLUDE | Yes, deterministic parser |
| DI Studio jobs → Workflows | Yes |
| SAS/STAT, IML → MLflow / Python | Yes |
| Row-level parity + exception report | Yes |
| Signed evidence pack, as-of lineage | Yes (Atlas) |
| Air-gapped / on-prem | Yes |
Already part-way through with Lakebridge? The certify pilot checks that output row by row instead of redoing it.
The first engagement
Fixed scope, fixed exit. You keep everything the pilot produces.
Want a smaller first step? The readiness scan inventories the estate in a week and is the smaller first step.
SAS Grid scales by adding more of the same. Databricks distributes the rewritten job across a cluster. The batch that finished at 6am now finishes before the trading desk opens.
A thousand %INCLUDE files is not a platform. The parser expands macros, then emits readable PySpark so a new hire can follow the match criteria.
SPD Server and .sas7bdat are not a lakehouse table. We land ACID tables with time travel so yesterday's extract is a version, not a copy job.
A classic SAS match-merge — two sorted datasets, BY key, IF IN= flags. Emitted as an explicit join, not a record loop.
/* SAS: customer + transactions */ proc sort data=cust; by cust_id; run; proc sort data=txn; by cust_id; run; data gold; merge cust (in=a) txn (in=b); by cust_id; if a and b; if amount > 1000 then segment = 'GOLD'; else segment = 'STD'; run;
# SAS match-merge → PySpark join
from pyspark.sql import functions as F
gold = (
cust.join(txn, "cust_id", "inner")
.withColumn(
"segment",
F.when(F.col("amount") > 1000, "GOLD")
.otherwise("STD"),
)
)
BY-group merge becomes an inner join. The IN= flags disappear because the join type is explicit. No sort required.
| SAS | Databricks | Notes |
|---|---|---|
| DATA step | PySpark DataFrame | Assignment, IF/THEN, BY-groups |
| PROC SQL | spark.sql / DataFrame | Joins, GROUP BY, CASE |
| Macro / %INCLUDE | Expanded then emitted | No leftover macro language |
| DI Studio job | Databricks Workflows | Replaces SAS Grid |
| SAS dataset / SPDS | Delta Lake | ACID + time travel |
| SAS/STAT | MLflow + Spark ML | Where a model actually ran |
A US federal agency, a UK public healthcare body, a US insurer, one of Canada's largest banks, FTSE 100 retail and energy companies, and a global marketing-data company. Customer names are shared under NDA in a demo, with reference calls on request.
See all engagements →SAS output compared to Databricks output — row by row, column by column. Differences flagged before sign-off, and the evidence lives on in Atlas after the migration ends. Atlas attaches to a conversion license or runs standalone.
See how Data Matching works → · What Atlas is →A SAS to Databricks pilot runs 4–6 weeks on about 10K lines of production SAS, validated row by row. A one-week readiness scan is the smaller first step.
Lakebridge is a free Databricks Labs toolkit for assessment, SQL conversion, and reconciliation, and you can keep using it for that. SAS2PY converts SAS DATA step, macros, DI Studio jobs, and SAS/STAT, and produces row-level parity with a signed evidence pack. If Lakebridge has already converted a workload, the certify pilot validates that output rather than repeating the work.
A representative SAS workload, agreed inputs with the original outputs (or a run in your environment so we can compare), and one named owner for weekly checkpoints. No code upload is needed to start — the intake form only asks what you are modernizing and when your SAS renews.
Three ways to start
Every route below ends with a person who has done this before, not a sales sequence.
Not ready to name a workload? The free sample assessment converts a representative piece of SAS at no cost — request one here.