Detect sensitive information and apply your team’s redaction and de-identification rules across documents, conversations, and datasets.
Step 01
Define and Align
Review synthetic documents, conversations, and records. Correct missed identifiers or over-redaction, and explain the transformations your policy requires.
Output schema
Deidentified textText
The source text or serialized record after the supplied redaction and generalization rules are applied.
Redacted typesChoose any
Identifier types replaced under the supplied policy; empty when none.
person_name
email_address
phone_number
customer_identifier
Generalized fieldsList of values
Fields reduced to a less specific value under the supplied policy; empty when none.
Needs reviewYes / No
Whether missing or ambiguous evidence prevents applying a required transformation.
Apply the supplied redaction and generalization policy to the source. Preserve permitted facts and flag unresolved transformations for review.
Validation accuracy—
Example 1 of 3
An email left in the document
Model input
Policy
Synthetic example. Replace individual names with [PERSON] and email addresses with [EMAIL]. Preserve organization names and non-identifying quantities. Do not add or infer personal information. Flag unresolved transformations for review.
Sutro can now generalize the rules based on your feedback.
Step 02
Model Sweep
Compare models on the same annotations and choose the quality–cost tradeoff for your workflow.
Compare models on expert agreement and cost. The interactive chart uses illustrative measurements.
Illustrative sweep · select Value or Best to use that model in your function.
Step 03
Deployment
Call Sutro Functions like any other software. Swap the model as soon as something more performant is released.
Selected modelWaiting for sweep results.
Real-time
import sutro as so
input_data = {
"policy": "Synthetic example. Replace individual names with [PERSON] and email addresses with [EMAIL]. Preserve organization names and non-identifying quantities. Do not add or infer personal information. Flag unresolved transformations for review.",
"source_type": "document",
"source": "Harbor Labs\nContact: Mira Patel\nEmail: [email protected]\nSeats requested: 12."
}
result = so.run_function(
"de-identification-pii-redaction",
input_data,
)
curl--fail-with-body-XPOST"${SUTRO_API_URL%/}/v1/functions/de-identification-pii-redaction/run" \
-H"Authorization: Key $SUTRO_API_KEY" \
-H"Content-Type: application/json" \
-d'{
"input": {
"policy": "Synthetic example. Replace individual names with [PERSON] and email addresses with [EMAIL]. Preserve organization names and non-identifying quantities. Do not add or infer personal information. Flag unresolved transformations for review.",
"source_type": "document",
"source": "Harbor Labs\nContact: Mira Patel\nEmail: [email protected]\nSeats requested: 12."
}
}'
Batch
import sutro as so
job_id = so.batch_run_function(
name="de-identification-pii-redaction",
data="deidentification_pii_redaction.csv",
)
results = so.await_job_completion(job_id)
curl--fail-with-body-XPOST"${SUTRO_API_URL%/}/v1/batch-inference" \
-H"Authorization: Key $SUTRO_API_KEY" \
-H"Content-Type: application/json" \
-d'{
"model": "de-identification-pii-redaction",
"inputs": [
{
"policy": "Synthetic example. Replace individual names with [PERSON] and email addresses with [EMAIL]. Preserve organization names and non-identifying quantities. Do not add or infer personal information. Flag unresolved transformations for review.",
"source_type": "document",
"source": "Harbor Labs\nContact: Mira Patel\nEmail: [email protected]\nSeats requested: 12."
}
]
}'
Apply your own transformation policy
Choose which identifiers to redact, which fields to generalize, and which facts to preserve. Teach those distinctions with reviewed examples, including shortened names and field-specific rules. These synthetic walkthroughs show document, conversation, and structured-record transformations under explicitly supplied policies.