Federate live data intoone governed view
Stop copying data between systems. Define a virtual view once, push down predicates to the right engine, and publish a governed surface to SQL, REST and AI agents — without moving a byte out of your environment.
// sources
// join key
// policies
// view
customer_360
rps
1.2k
p95
87 ms
cache
94%
One layer for federated truth
Virtualize is the entry point of every governed data product on SOFI — federate, govern and cache in one definition.
Federate live sources
Push down predicates to Postgres, Oracle, MySQL, MongoDB and more. SQLGlot rewrites — DuckDB stitches.
Govern as you model
Masking, RBAC and lineage are part of the view definition — not a wrapper layered on top.
Cache without surprises
Choose live, cached or hybrid per view. Invalidations honor source freshness and consumer SLAs.
From scattered sources to a published view
Four steps. One definition. Zero data movement.
Connect
postgres.core · oracle.billing · mongo.events
Model
customer_360 view · 14 columns · 3 sources
Govern
pii.masked · audit.all · rbac.tenant
Publish
sql · rest · mcp · jdbc
Pick the language that ships your stack
Same definition, four ways to drive it. Python, Node, SQL or REST — they all hit the same governed endpoint.
from sofi import Sofi
sofi = Sofi(api_key="YOUR_KEY")
view = sofi.virtualize(
name="customer_360",
sources=[
"postgres.core.customers",
"oracle.billing.invoices",
"mongo.events.sessions",
],
join_on="customer_id",
policy=["pii.masked", "audit.all"],
)
view.publish(channels=["sql", "rest", "mcp"])Patterns that ship in under two weeks
Virtualize is the foundation under every governed data product on SOFI.
// pattern
Customer 360°
Stitch CRM, billing and product events into a single governed entity. Masked PII, fresh metrics, no warehouse round-trip.
// pattern
Real-time analytics
Query operational data without ETL. Cache hot windows, push down filters, leave the source of truth in place.
// pattern
AI grounding
Expose governed views as MCP tools. Agents pull fresh enterprise data with the same RBAC humans get.
// pattern
DB migration
Decouple apps from physical sources during a cutover. Swap engines underneath while consumers see one stable view.
Built for production not benchmarks
Predicate pushdown, cache, and policy evaluation in one controlled query path.
<800 ms
federated query p95
Predicate pushdown plus DuckDB-orchestrated joins keep cross-source latency in BI-friendly territory.
94%
cache hit rate
Hot windows are pinned in Redis. Stale-while-revalidate keeps freshness honest without a stampede.
0
bytes leaving the perimeter
Sources stay in your private network. SOFI federates inside the perimeter and publishes governed views outward.
10+
engines federated
Postgres, MySQL, SQL Server, Oracle, MongoDB, ClickHouse, Cassandra, Neo4j, Elastic — plus CSV/JSON/Parquet.
Questions about virtualize
What enterprise architects ask before adopting Virtualize as the federation layer.
No. SOFI keeps source data in place and federates queries on demand. Cached responses are governed by per-view freshness rules and stay inside your environment.
// ready to virtualize
Ship your first governed view this week.
Connect three sources, model a view, attach a policy and publish. Trial includes private deployment and unlimited views.