~200
Submitting nations, one schema
at a multinational financial institution
Fleak joins the Databricks startup accelerator. See the announcement
Regulated data, normalized inside your perimeter.
Fleak normalizes security logs, transaction feeds, and regulatory submissions to one schema in real time, masks sensitive values in flight, and routes each record to the system that needs it. It runs on-prem or air-gapped, and every transformation is logged.
- Splunk
- Microsoft Sentinel
- Kafka
- Snowflake
- Databricks
- On-prem & air-gapped
30–50%
SOC spend reduction
every log routed to where its job gets done
0
AI inference on production data
the Brain sees samples, never live records
Every system speaks its own format.
Regulators expect one answer.
Core banking, card processors, trading systems, SOC tools, and counterparties describe the same activity differently. Risk and compliance teams reconcile them by hand.
Same label, different meaning
A transaction date in one feed is a settlement date in another. Matching on field names gets the labels right and the meaning wrong.
Data that can't leave the building
Residency rules and internal policy keep raw customer data on-prem. Cloud-only pipelines and AI tools that need raw records are off the table.
Every derived figure is evidence
Auditors ask how each number was produced. Pipelines built from scripts and spreadsheets can't show their work.
Semantic normalization,
with the audit trail built in.
The Brain resolves what each field means, not just what it is called. The Muscle applies the approved mapping deterministically, and every transformation is logged.
Meaning, not just labels
The Brain maps fields by what they mean, so a settlement date and a transaction date stay distinct even when two sources name them the same.
Every source as it arrives
APIs, message queues, files, and legacy exports normalize to one canonical schema. Upstream teams and counterparties keep sending their native format.
Security logs in OCSF
Identity, endpoint, and network telemetry lands in OCSF, so detections and investigations run against one schema.
Sensitive values masked
Card numbers, account numbers, SSNs, and emails are masked, hashed, or encrypted in flight, before any destination receives the record.
Inside your perimeter
Run Fleak in your cloud, on-prem, or air-gapped. Production data is processed in memory and never sent to an AI model.
An audit trail on every change
Every transformation and routing decision is logged and attributable, and every config change is reviewed before it deploys. Built for DORA, SOX, and PCI DSS evidence.
Lineage for regulators
Trace every reported figure back to its source record and the mapping version that produced it, for BCBS 239 and DORA reviews.
Retention without SIEM prices
Keep every log for its mandated retention period in your own storage, and pay SIEM prices only for what detections use.
Residency by region
Process and store regulated data in the region it belongs to, with policies per jurisdiction.
Normalize inside your perimeter.
On-prem or air-gapped. Every transformation logged. No production record ever sent to an AI model.
The Brain plans.
The Muscle runs.
Fleak resolves each source against one canonical schema, then routes by purpose. Fraud models get real-time signals. The SOC gets OCSF. Compliance gets an archive with lineage. The Brain plans the mapping once, and the Muscle executes it deterministically.
Built for
- Retail & commercial banks
- Payments & cards
- Capital markets
- Insurance
- Regulators & multilaterals
A cross-border payment arrives in three formats.
Where does each record need to go?
Fraud detection
Real timeNormalized transaction events stream to fraud models as they happen, with every field resolved to one meaning.
SOC & SIEM
OCSFAuthentication and access events from the same systems land in OCSF for detection and investigation.
Regulatory archive
AuditFull-fidelity records go to retention storage with the logged transformations that produced every derived field.
A multinational financial institution maps submissions from nearly 200 member nations to one schema this way, without asking any of them to reformat.
"Meet every source where it is. The Brain resolves meaning, the Muscle executes deterministically, and nobody upstream has to reformat."
Financial services · Fleak Platform
Bring a feed your teams reconcile by hand.
30-minute working session. We'll normalize it live and show you the mapping.
Explore Related
Every source in OCSF, long tail included
211 templates, plus AI-generated mappings for the rest.
See Detail →Mask sensitive data in flight
Detectors, custom patterns, hashing, and quarantine.
See Detail →Run both SIEMs, cut over by source
Every source dual-shipped from one pipeline.
See Detail →