Subtl.ai: Building the 'Private Google' for the Global Banking and Enterprise Sector
In the massive vault of a modern bank, the most valuable asset isn't just the gold or the cash—it's the data. Yet, when an officer needs to verify a specific RBI regulation or a complex legal contract across 3,000 documents, the current solution is often a primitive "Control-F" or a long wait for an expert's time. Vishnu Ramesh, Co-founder and CEO of Subtl.ai, is building the bridge to instant answers. By creating a "Managed Search Service" that understands intent rather than just keywords, Vishnu is helping enterprises build their own private search engines with human-level accuracy.
Vishnu Ramesh’s journey with Subtl.ai is a story of academic depth meeting entrepreneurial grit. An alumnus of BITS Pilani Hyderabad, Vishnu started as a summer intern in 2018 building a tech prototype. Under the guidance of Chief Scientist Prof. Manish Shrivastava, that internship evolved into a deep-tech venture that secured a landmark contract with the **State Bank of India (SBI)**. Today, Subtl.ai is proving that its RAG (Retrieval Augmented Generation) models can outperform global giants when it comes to the nuances of complex, domain-specific regulations.
The Information Overhead
Enterprises today suffer from "Information Overhead." Employees spend hours opening PDFs one by one or waiting for experts to answer simple queries. Subtl.ai solves this by letting users ask natural language questions—the same way they would ask a human expert—and receiving a direct, highlighted answer from their private cloud.
The Problem: When "Control-F" Is Not Enough
Traditional search engines rely on keywords. If you search for "interest rates," you get every document that mentions those two words. But in banking, a user might ask: "What is the penalty for early withdrawal under the new liquidity guidelines?" A keyword search fails here because it doesn't understand the *intent* of the question. For high-stakes industries like BFSI (Banking, Financial Services, and Insurance), a 90% search accuracy is the difference between a compliant transaction and a massive regulatory fine.
"pdf files are a nightmare for data engines," Vishnu explains. "They have paragraph breaks, tables, and images scattered in a non-linear way. Our first challenge was building a pillar of tech that could cleanly extract this data so it could be indexed for a smart engine."
The Solution: A Domain-Focused Search Engine
Subtl.ai is more than just a search bar; it is a managed search service built on three pillars of technology. It ignores keywords entirely, focusing instead on the **Linguistic Ontology** of a specific domain—meaning it understands the jargon of a banker differently than the jargon of a lawyer.
The Subtl.ai Intelligence Flow
- High-Fidelity Extraction: Cleaning and structuring complex data from PDFs, tables, and paragraphs.
- Intent-Based Mapping: Using AI to understand the *meaning* of a user's question, regardless of the words used.
- Relevance Scoring: Every answer is given a color-coded confidence score (Green for high, Red for suspicious).
- Highlighted Fulfillment: The engine doesn't just give a link; it opens the document and highlights the exact answer in a new tab.
Achieving 93% Accuracy in Banking
Subtl.ai’s "North Star" is hitting a quarter-million users, a target they are rapidly approaching through their traction in the banking sector. In internal tests, their models achieved **93% accuracy** in the banking domain, significantly outperforming generic open-source models on over 51,000 RBI regulations.
"Track record matters in BFSI," Vishnu notes. "Banks need to be 100% sure that the solution will solve their problem. We've fixed the 'drift' between what users want and what is being offered by building models that learn the specific linguistics of the industry."
Subtl.ai Scale & Performance
- 93% Accuracy: Peak performance achieved in the banking sector for complex regulatory queries.
- 250,000 User Target: Scaling rapidly toward a quarter-million users across enterprise clients.
- API-First Strategy: Integrating with AML and CRM platforms to provide search as a "Value-Added Service."
- On-Premise Security: Deploying directly on bank servers to ensure private data remains 100% secure.
The Meaning of Entrepreneurship: Mental Wrestling
For Vishnu, building a startup is an exercise in "mental wrestling." It involves challenging every member of the team to do more while maintaining a bi-directional respect. He believes that a founder's role is to transition from being the person who builds the prototype to the person who builds the engine of growth.
"You cannot do everything on your own," he admits. "You have to delegate. My co-founder handles the product initiatives, while I oversee outreach and operations. But we still wrestle with ideas in the drawing room to ensure our tech remains world-class."
Founder's Lesson: The Window of Opportunity
Vishnu compares startup success to an artist becoming "famous overnight." It only looks like it happened quickly because people ignore the five to ten years of consistent work that came before the window opened. Be ready to write out the rejections until your window arrives.
The Future: Beyond Text to Databases
Subtl.ai is already looking at the next frontier: **Structured Data**. While they have mastered the unstructured world of PDFs, they are exploring how to apply the same intent-based search to massive databases and tables. Their vision is to create a truly customized information gateway where every employee can find any piece of enterprise knowledge instantly.
Subtl.ai is proving that in the age of information overload, the most powerful tool isn't the one that gives you the most data—it's the one that gives you the exact answer. By turning "Control-F" into "Intent-Search," Vishnu Ramesh and his team are ensuring that for the modern enterprise, no answer is ever out of reach.
About the Guest
Vishnu Ramesh is the Co-founder and CEO of Subtl.ai. An alumnus of BITS Pilani Hyderabad, he is a deep-tech innovator with a focus on cognitive search and natural language processing. Under his leadership, Subtl.ai has become a key player in the enterprise search space, securing landmark contracts with major banks and defense PSUs. He is a passionate advocate for the use of vertical AI to solvebackend operational gaps in high-stakes industries.
Subtl.ai is an AI-powered managed search platform that turns enterprise documents into a private search engine. Based in Hyderabad and partnered with IIIT Hyderabad, the company specializes in intent-based information retrieval for the BFSI, manufacturing, and legal sectors, helping organizations reduce information overhead and improve operational efficiency.