Bring your own Custom LLM Models in UiPath: Revolutionizing Feedback and Survey Analysis

Bring your own Custom LLM Models in UiPath: Revolutionizing Feedback and Survey Analysis

Use Case Description

Dearest gentle Readers,

In today’s data-driven world, businesses thrive on insights derived from customer feedback, surveys, and reviews. The ability to quickly and accurately categorize this data can significantly impact product and service improvements. Leveraging custom Large Language Models (LLMs) and integrating them with UiPath’s robust automation platform opens new horizons for intelligent automation. In this blog, I’ll walk you through how I built a custom LLM model, deployed it using UiPath’s AI Center, and seamlessly integrated it into UiPath Studio to transform feedback analysis—all within the Automation Suite! :grin:
The Value of Customer Feedback
Customer feedback, surveys, and reviews are gold mines of information for any business. However, manually sifting through this data to extract meaningful insights is labor-intensive and prone to errors. By automating the categorization of this data using custom LLM models, businesses can:
• Quickly identify key themes and trends.
• Pinpoint specific areas for product or service improvement.
• Enhance customer satisfaction through timely responses to feedback.
Building the Custom LLM Model
Creating a powerful custom LLM model starts with the right dataset. I gathered a diverse set of feedback and survey data to train the model, ensuring it could accurately recognize and categorize various themes and sentiments. The training process involved:

  1. Data Preparation: Cleaning and preprocessing the data to make it suitable for training.

  2. Model Selection: Choosing an appropriate LLM architecture that supports our categorization tasks.

  3. Training: Running multiple iterations to fine-tune the model for optimal performance.

  4. Evaluation: Testing the model on a separate dataset to validate its accuracy and robustness. Challenges such as handling noisy data and ensuring the model’s generalizability were addressed through rigorous validation and fine-tuning.
    Deploying the Model with UiPath AI Center
    Deploying the model with UiPath AI Center was a pivotal step. The model, based on LLM architecture, processed unstructured text and provided JSON output with entity details and corresponding values. Here’s how I did it:

  5. Model Packaging: I packaged the model into a zip file compatible with UiPath AI Center. The main components required were the main.py file (which contained the Python wrapper code to call the model), requirements.txt, and on-prem dependencies folder which included all the python libraries mentioned in the requirements.txt compatible with the environment -this step can be skipped if you are on UiPath cloud. Its an important additional step for on prem automation suite setup.

  6. Uploading to AI Center: I zipped these components and uploaded them using the AI Center’s upload zip file option. I then created a skill in AI Center.
    Integrating the Model in UiPath Studio
    Integrating the deployed model into UiPath Studio was straightforward:

  7. Skill Invocation: I called the AI skill directly from UiPath Studio using the ML Skill activity

  8. Data Processing: The robot processed feedback and survey data, passing it to the custom LLM model for categorization.

  9. Insight Generation: The output from the model was analyzed to generate actionable insights.
    This seamless integration allows for real-time analysis and decision-making, empowering businesses to act swiftly on customer feedback.
    Conclusion
    Integrating custom LLM models with UiPath offers numerous benefits:
    • Efficiency: Automates tedious data categorization tasks, freeing up valuable time for employees.
    • Accuracy: Provides consistent and accurate insights, reducing human error.
    • Scalability: Easily scales to handle large volumes of feedback and surveys.
    By harnessing the power of custom LLM models and UiPath’s AI Center, businesses can unlock deeper insights from their feedback and surveys, driving continuous improvement.

I hope this post inspires you to explore the incredible potential of AI in automation. Share your experiences and join the conversation on pushing the boundaries of intelligent automation!

AS-IS WORKFLOW, TO-BE WORKFLOW

-

Other information about the use case

Industry categories for this use case: BFSI, Compliance, Customer Service, Finance, Healthcare Pharma, HR, Information Technology and Services, Legal, Logistic, Manufacturing, Marketing Sales, Operations, Telecom, Universities Academy, Banking, Insurance, Public Sector, Other Sector

Skill level required: Advanced

UiPath Products that were used: UiPath AI Center

Other applications that were used: -

Other resources: -

What is the top ROI driver for this use case?: Accelerate growth and operational efficiency

7 Likes

Hi @vaishnavi_suresh i am working on UiPath document understanding ml model but i dont know how we can create custom model and upload into ai center cqn you pleasw share your inputs it would be really helpful for me .

Thank you

For building custom models, you can refer to the detailed documentation here: Docs: Building ML Packages.

Additionally, the AI Center offers a wide range of pre-built models in the Out-of-the-Box section. However, hosting a custom LLM model is an excellent approach for @vaishnavi_suresh to explore—well done on making that happen! :ok_hand: