Gi-Net
Enabling GenAI-driven operations in Telecom.Gi-Net is a framework that provides low code integration to enable all your legacy Telecom OSS/BSS applications to be GenAI capable.
Capabilities
Traceability
Detailed transaction management system
Ability to track query, response, reasoning, and source references
Data Privacy
Full control over training data
No use of customer data for training
AI Governance module for control and enhancement of LLM responses
Features
Continuous Learning:
learns from user feedback
Governance module:
monitors, manages, and approves user feedback
2
TSLAM:
7B parameters, 16-bit quantized for local deployment
3
1
Industry’s first telecom-specific LLM
TSLAM-7B
Intent-to-API conversion:
converts user intent to network-specific APIs
Chat Interface
get details on uploaded docs and integrated applications
5
No code Interface
easily integrate legacy applications and documents
6
4
The Future of Telecom is AI-Driven
NetoAI’s Gi-Net (Generative Intelligent Network) framework seamlessly integrates TSLAM, empowering telecom operators to achieve 100% optimization without disrupting their existing infrastructure.
With Gi-Net and TSLAM, telecom operators can:
- Reduce operational costs: By automating tasks and minimizing downtime.
- Improve customer satisfaction: By ensuring reliable and efficient network performance.
- Gain a competitive edge: By leveraging AI-powered insights and optimizing network operations for maximum efficiency.
Gi-Net has a continuous learning module through which it learns and enhances itself. Users can also provide feedback to the framework to enhance the responses as required. Additionally, an AI governance module is in place to monitor and manage user feedback to enable LLM to enhance itself. Please find screenshots from Gi-Net below,
Capabilities of Gi-Net
Gi-Net’s TSLAM (Telecom Specific Large Action Model) is finetuned with a blend of general purpose and as well telecom-specific data (non-copyright protected data such as open-source general datasets, IETF RFCs, and configuration guides). This fine-tuning aims to equip the model with a deep understanding of telecom-specific activities, enabling it to effectively interpret and respond to telecom-specific queries.
Gi-Net provides two interfaces listed below,
- Document Interface:
- An interface that facilitates the uploading of any documentation and provides a chat interface for users to query data related to the documents.
- Application interface:
- Seamlessly integrate existing applications using a low-code approach.
- Enables the existing apps to be GenAI capable, providing a chat interface for users to interact with the applications and perform various actions.
- Enables Gi-Net to take intelligent actions via the apps.
- No code changes are to be done on the existing applications.
Gi-Net has a continuous learning module through which it learns and enhances itself. Users can also provide feedback to the framework to enhance the responses as required. Additionally, an AI governance module is in place to monitor and manage user feedback to enable LLM to enhance itself.
Traceability of Gi-Net
One significant drawback/limitation of all GenAI models is their inability to trace and explain their responses. In the Telecom domain, it’s crucial to understand why a model has made a particular decision, as allowing AI to autonomously manage network and operations (without oversight) is not feasible.
Gi-Net comes with a very detailed transaction management system through which you’ll be able to identify,
-
- What’s the query?
- What’s the response?
- Why this response?
- What’s the source referenced for this response?
If dissatisfied with a response, you can provide negative feedback to the model and rectify it via the continuous learning module.
This ensures that Gi-Net considers your input and delivers accurate responses for similar queries in the future.
Data privacy on Gi-Net
All your data used for training and inference by Gi-Net will be under your full control.
The knowledge graph created by Gi-Net and TSLAM-7B with your data can be deployed either on your private cloud or on-prem. You get to decide who can control and manage the data.
After deployment, NetoAI doesn’t require any access to your environment unless there’s a specific need to perform any enhancements.
All data loading, feedback for continuous learning, and integration with application data remain under your control. We strictly adhere to privacy and confidentiality standards and do not utilize or access any of your data for model enhancement purposes.