NIN-CallLabels-v1
Label English bank-support transcripts for banking intent, UPI payment issues, refund requests, agent promises and resolution status. Designed for bank operations and complaint review.
Explore banking model ↗Focused models.
Built for real work.
Upcoming private AI models for enterprises, government and banks. Specialized workflows, supported by focused privacy, speech and text models.
Explore upcoming modelsAll models are upcoming. Capabilities and deployment targets describe our plans.
Upcoming
PRIVACY / PII REDACTION
NIN-Redact-v1
Share the context.
Keep personal details private.
Planned detection and masking of personal information in notes, documents and transcripts before sharing. For private work in NINXL and workflows that use external models.
Explore model ↗REDACT / BEFORE SHARINGIllustrative example
Original text · fictional details“Please send Maya’s meeting notes to maya@example.com. You can reach her at +91 90000 00000.”
Redacted text“Please send [PERSON]’s meeting notes to [EMAIL]. You can reach her at [PHONE].”
- Person
- [PERSON]
- Email address
- [EMAIL]
- Phone number
- [PHONE]
Manually illustrated output, not a live model prediction. Detection coverage will need evaluation before release.
SPECIALIZED MODELS
Built for institutional work.
Domain-specific extraction and labels for the people reviewing cases, applications and conversations.
NIN-CaseExtract-v1
Extract reported issues, customer requests, commitments and follow-ups from enterprise support transcripts and service tickets for case-management review.
Explore enterprise support model ↗NIN-GovExtract-v1
Extract application references, requested services, submitted document lists and grievance details from government text records for officer review and administrative workflows.
Explore government model ↗THE MODEL FAMILY
A focused job for every model.
13 upcoming models: specialized models for banking, government and enterprise support, with shared privacy, speech, text and action capabilities.
01 / PrivacyUpcoming
NIN-Redact-v1
Keep personal details out of shared text.
Detect and mask PII in notes, transcripts and documents before sharing or passing text to another model. Planned coverage includes Indian identifiers.
On-device / private-hosted · plannedExplore model ↗
02 / SpeechUpcoming
NIN-Transcribe-v1
Bring your voice into text.
Transcribe dictation, meetings and recordings for NINK and NINXL. Indian-language support is planned.
On-device / private-hosted · plannedExplore model ↗
03 / TextUpcoming
NIN-Extract-v1
Turn everyday text into structured fields.
Extract specified fields from documents, messages and notes for forms, records and workflows.
On-device / private-hosted · plannedExplore model ↗
04 / TextUpcoming
NIN-Tag-v1
Give your notes and files useful labels.
Suggest topics and categories for notes, documents and conversations to help organize and retrieve them in NINXL.
On-device / private-hosted · plannedExplore model ↗
05 / ActionsUpcoming
NIN-Action-v1
Turn intent into a focused action.
Select and execute bounded tool actions within NINXL.
Local NINXL · plannedExplore model ↗
06 / ActionsUpcoming
NIN-Route-v1
Give each request a suitable path.
Route requests to suitable tools and workflows.
Local NINXL · plannedExplore model ↗
07 / TextUpcoming
NIN-CallSummary-v1
Keep the important parts of a conversation.
Produce transcript-grounded summaries of meetings and support calls, including decisions and follow-ups.
On-device / private-hosted · plannedExplore model ↗
08 / BankingUpcoming
NIN-CallLabels-v1
Turn bank-support conversations into reviewable labels.
Label English bank-support transcripts for banking intent, UPI payment issues, refund requests, agent promises and resolution status. Designed for bank operations and complaint review.
On-device / private-hosted · plannedExplore model ↗
09 / Enterprise supportUpcoming
NIN-CaseExtract-v1
Prepare enterprise support cases for review.
Extract reported issues, customer requests, commitments and follow-ups from enterprise support transcripts and service tickets for case-management review.
On-device / private-hosted · plannedExplore model ↗
10 / GovernmentUpcoming
NIN-GovExtract-v1
Structure government applications and citizen requests.
Extract application references, requested services, submitted document lists and grievance details from government text records for officer review and administrative workflows.
On-device / private-hosted · plannedExplore model ↗
11 / TextUpcoming
NIN-Language-v1
Recognize the language before the next step.
Identify language and mixed-language content in text to choose suitable rewrite and workflow paths. Indian languages are part of the roadmap.
On-device / private-hosted · plannedExplore model ↗
12 / SpeechUpcoming
NIN-Align-v1
Connect each word to its moment.
Align transcript words with recording timestamps for captions, playback navigation and searchable meeting notes.
On-device / private-hosted · plannedExplore model ↗
13 / SpeechUpcoming
NIN-ClearSpeech-v1
Bring the voice forward.
Reduce background noise in voice recordings before dictation and transcription for NINK and NINXL.
On-device / private-hosted · plannedExplore model ↗
WHERE THE WORK HAPPENS
Close to your data.
On your device.
Local execution is the goal for NIN’s model family, with integrations planned for NINK and NINXL. Supported devices and model requirements will be defined as each model is evaluated.
Inside your organisation.
Private-hosted deployment is planned for teams that need models within their own environment, including support operations and banking workflows.
WHAT TO EXPECT
A clear starting point.
Who are these models for?
Enterprises, government departments and banks building private AI workflows. Specialized models address banking complaints, enterprise support cases and government records; shared models support privacy, speech and text processing across these environments. Integrations with NINK and NINXL are also planned.
How are the models intended to work together?
A workflow could transcribe a meeting, summarize it, extract follow-ups and tag the notes. Redaction could mask personal details before sharing. Action and routing models are planned to help select suitable tools within NINXL.
Will PII redaction cover Indian identifiers?
Indian identifiers are part of the planned coverage. Exact supported entity types, detection accuracy and limitations will be documented after evaluation. This roadmap does not promise complete detection.
Can I download or integrate a model today?
These models are upcoming. Downloads, supported platforms, integration instructions and benchmarks will be published when available. Contact NIN to discuss a workflow or an evaluation partnership.
Bring the work you want to simplify.
Planning a banking, government or enterprise workflow? Tell us about your review process and private deployment requirements.
Discuss an institutional pilot