Introducing GenAI Features in Catalyst QuickML

Introducing GenAI Features in Catalyst QuickML

Hi everyone,

Building machine learning models can often feel slow and complex, especially when teams wait for perfect certainty before testing their ideas. But in reality, faster progress comes from early experimentation—trying out models quickly, learning from results, and iterating along the way. That’s where Catalyst QuickML comes in.

What is Catalyst QuickML?

QuickML is a low-code cloud platform designed to help you build, train, and deploy machine learning models, without needing to write complex code or manage infrastructure. It utilizes a drag-and-drop pipeline builder, allowing you to organize your data and machine learning operations into configurable nodes. With just a basic understanding of machine learning and data preprocessing, you can quickly go from raw data to a working model and deploy it in your application instantly using the generated endpoint URL.

QuickML is designed to be both cost-effective and secure, making it a practical choice for businesses of all sizes. It ensures secure data handling aligned with industry standards, runs on a cloud-based infrastructure that eliminates the need for expensive hardware, and tracks compute usage during model training and API calls. With a transparent pay-per-use billing model, QuickML allows teams to scale their AI efforts while keeping costs under control.

Whether you're an analyst, product manager, or data enthusiast, QuickML's clean interface and guided workflows make machine learning approachable and usable. It helps you stay focused on solving business problems without complex frameworks or cloud setups.  

New GenAI features now available in QuickML

QuickML already comes equipped with essential features like automatic dataset profiling upon import, powerful visualization tools, a drag-and-drop pipeline builder, model evaluation, AutoML, and more. You can explore these core capabilities in detail through our help documentations.

Now, QuickML is becoming even more capable with the introduction of a new set of GenAI features that bring natural language understanding and knowledge-based intelligence directly into your workflows.
Info
The new GenAI capabilities are currently available in US, EU, and IN data centers.
Let’s take a closer look at what’s new.

LLM Serving: Flexible, real-time AI conversations

QuickML now makes it simple to experiment with and deploy large language models (LLMs) through an intuitive chat-based interface. These models generate responses based on patterns learned from vast amounts of pre-trained data, including books, websites, and articles.
 
What makes it different?

Unlike many platforms that offer limited options to tweak, QuickML gives you fine-grained control without the complexity.
 
1) Tune the tone and output — Control creativity, coherence, and response length directly in the UI.

2) Switch between models — Test multiple LLM models such as Qwen 2.5 - 14B Instruct, Qwen 7B Coder, and Qwen 2.5 - 7B Vision Language Model in the same chat to see which suits your use case best. Support for many more models will be added in the future.

3) Parameter configuration — While exploring LLM models in the chat interface, you can fine-tune parameters to verify whether the model produces the desired results.

4) Instant deployment — Use the auto-generated endpoint URL to embed your LLM in third-party tools or customer-facing apps.

5) Built for everyone — No need for complex infrastructure or heavy engineering, even non-technical teams can get started in minutes.

Example: Let’s say your support team wants to create an internal assistant to handle employee queries around policies or benefits. They can choose a base model, adjust the tone, test outputs, and deploy the assistant to their business application such as Slack, all without writing code or managing servers.
 

 

RAG (Retrieval-Augmented Generation): Grounded AI you can trust

While LLMs are powerful, QuickML’s RAG feature enhances their effectiveness by connecting your models to your actual content and ensuring transparency in how answers are generated.
 
Here’s how RAG in QuickML stand out:
 
1) Document-aware responses — Connect your model to internal knowledge bases, which consists of help docs, FAQs, and reports.

2) Deep Zoho integration — Seamlessly pull documents from Zoho WorkDrive and Zoho Learn to enrich your AI’s knowledge.

3) Transparent breakdowns — Each response includes a detailed view of which documents were used and which parts contributed to the answer, so you know exactly where the model’s reasoning comes from.

4) Secure and controlled — All data stays within your control; it's accurate with no pre-trained answers from the model.

5) Instant deployment — Use the auto-generated endpoint URL to embed your RAG in third-party tools or customer-facing apps.

Example: Suppose your legal or compliance team wants to enable employees to query policy documents. With RAG in QuickML, they can build a chat assistant that only responds based on verified, internal documents with full traceability.

Likewise, if your customer support team receives frequent questions about product features, troubleshooting steps, software error messages, or return policies, they can use RAG in QuickML to create an AI assistant trained on support documentation, help center articles, and internal SOPs. The assistant will provide accurate, reference-backed responses, reducing support load and improving response consistency without risking misinformation.


Chart Insights: Understand your data in plain language

Charts can show trends, but understanding why those trends matter is quite important, especially when you're juggling multiple datasets or presenting results to non-technical stakeholders. That’s where Chart Insights in QuickML helps.

As you build visualizations in QuickML, you’ll now see a Chart Insights icon next to each chart. Clicking it opens a side panel that presents natural-language explanations about the chart, describing patterns, anomalies, distributions, and other meaningful observations.
 
You can access Chart Insights in two places:

1) During chart creation — When generating a chart, click Generate Key Insights to instantly see what the model observes in the data.

2) After chart creation — In the Visualization section, you can revisit and explore insights for any saved chart.



Idea
Sneak Peek: Upcoming Enhancements in QuickML's GenAI Features

We’re excited to share that two major enhancements will soon be available in QuickML's LLM Serving feature:

1. Two Interaction Modes for LLM Serving

One-Shot Interaction Mode – The model responds only to the current prompt, without referencing earlier messages.
Conversation Mode – The model maintains context across multiple prompts to generate relevant answers in ongoing conversation.

2. LLM Endpoints & Parameter Configuration Saving

 While you can already adjust parameters such as temperature, max tokens, and Top-k, the upcoming update will let you save those parameter configurations and generate unique endpoint URLs tied to them. This means you can reuse configurations and integrate endpoints that exactly match your business needs.

These features mark a major step in QuickML’s journey, making advanced AI tools usable, explainable, and efficient for everyday business processes. For more information on these features, please refer to the help documents on LLM Serving, RAG, and Chart Insights available on our help resources page.

If you haven’t explored Catalyst QuickML yet, now’s the ideal time to explore. Try out the GenAI features and let us know your thoughts. We’d love to hear your feedback and suggestions in the comments.
 
Have questions or need assistance? Feel free to reach out to our support team at support@zohocatalyst.com. We’re here to help.
 
Thanks, and have a great day!
 
Regards,
Varsha P
Catalyst QuickML

    • Announcements

    • Introducing GenAI Features in Catalyst QuickML

      Hi everyone, Building machine learning models can often feel slow and complex, especially when teams wait for perfect certainty before testing their ideas. But in reality, faster progress comes from early experimentation—trying out models quickly, learning

    • Announcing Deprecation of Catalyst File Store, Event Listeners, and Cron

      We would like to announce that the following Catalyst features are now in their deprecation phase and will reach End Of Life (EOL) on 30 April, 2026- Catalyst File Store Catalyst Event Listeners Catalyst Cron New users who sign up for Catalyst from today

    • React Nexus 2025 Recap: Catalyst Slate in Action!

      Hey Catalyst Community! We recently attended the React Nexus 2025 conference, an exciting gathering for frontend enthusiasts and React developers. Our team had an incredible time presenting and conducting a hands-on workshop on Catalyst Slate, our streamlined

    • [Webinar] A hands-on guide to Catalyst Stratus

      Have you used Catalyst Stratus yet? It’s an object storage service that makes it easy to handle large files — whether they're coming from your Catalyst app or other Zoho apps. We’re hosting a live coding session where you’ll build a working prototype

    • Catalyst Video Tutorials!

      Hello everyone! We’ve been brewing something exciting behind the scenes, and we’re thrilled to finally share it with you- Catalyst video tutorials are here! We recognized that videos are the predominant medium for learning and discovery these days, so

      Catalyst Community