IIIT Hyderabad’s Subtl.ai V2: A Game-Changer in Enterprise AI
Published on 1/3/2025 | 4 min read
Published on 1/3/2025 | 4 min read
Vishnu Ramesh, the founder of Subtl.ai, recently shared significant advancements in the startup’s enterprise AI offerings on LinkedIn. Subtl V2 introduces enhanced features aimed at improving AI-driven solutions for private and sensitive data in enterprise applications. Developed in collaboration with engineers and researchers from IIIT Hyderabad, Subtl V2 is a powerful tool for businesses looking to harness AI securely and efficiently.
Custom Embedding Models: Subtl V2 introduces a new “co-create embeddings program,” allowing developers to design proprietary Retrieval-Augmented Generation (RAG) embedding models tailored specifically for private document datasets. This feature enables businesses to create customized AI solutions based on their unique data needs.
Enhanced Data Visualization: In the new version, Subtl.ai now includes Excel-like grid views for structured data, providing enterprises with better accessibility, deeper insights, and a more intuitive interface for managing large datasets.
No Limits: One of the most anticipated upgrades in Subtl V2 is the removal of restrictions on document size and page counts. This change ensures that enterprises can scale their operations seamlessly without the limitations that typically hinder the processing of large datasets.
Since its inception in 2020, Subtl.ai has made significant strides in the enterprise AI space, particularly by providing secure and efficient AI solutions for managing private and sensitive data. A key success story involves the State Bank of India (SBI), which adopted Subtl.ai for its internal operations. The results were impressive:
92% Accuracy: Subtl.ai achieved a high level of precision in information retrieval, greatly improving the efficiency of SBI’s internal processes.
Time Savings: The system saved a total of 56,570 minutes, which translated into financial savings of ₹5 lakh for SBI. This demonstrates the real-world value of Subtl.ai’s AI solutions in streamlining enterprise operations.
In a conversation with Analytics India Magazine, Ramesh elaborated on Subtl.ai's mission to reduce reliance on global AI giants like OpenAI. He explained the security concerns associated with shared endpoints in solutions offered by major companies like Microsoft. For organizations in sectors that handle sensitive data, such as banking, these security gaps make it difficult to trust shared AI solutions.
Positioning Subtl.ai as a “private Perplexity,” the startup offers localized, secure solutions for enterprises. Built on the Llama 3 8B variant, Subtl.ai allows organizations to:
The enterprise AI space is becoming increasingly competitive, with established players like Google, Anthropic, and OpenAI offering robust retrieval capabilities through their APIs. Additionally, Perplexity has introduced its own premium offering, "Internal Knowledge Search," targeting enterprises.
However, concerns around data privacy and security continue to be a significant barrier for many enterprises. Incidents such as the Amazon Q data leak have highlighted the risks associated with cloud-based data processing, giving localized, secure solutions like Subtl.ai a significant competitive edge. Enterprises in regulated industries, such as banking, healthcare, and government, are particularly sensitive to these risks and are seeking AI tools that can ensure strict data privacy.
Subtl.ai’s focus on providing secure, private AI solutions has positioned it as a leader in the enterprise information retrieval market. By leveraging advanced embedding techniques and emphasizing secure, localized data management, Subtl.ai is reshaping the way businesses manage and retrieve critical information. As the demand for privacy-conscious AI solutions continues to grow, Subtl.ai is poised to play a pivotal role in the evolution of enterprise AI.