Extract ‘relevant info’ with Subtl.ai’s new tech

Hyderabad: Subtl.ai, a Hyderabad-based deep tech startup, has developed a technology that can read through unstructured text documents and give users relevant information. This will come in handy for banks and educational institutions where similar queries repeat. The technology, which will be put to use through a bot, will search information repositories and give the […]

Hyderabad: Subtl.ai, a Hyderabad-based deep tech startup, has developed a technology that can read through unstructured text documents and give users relevant information. This will come in handy for banks and educational institutions where similar queries repeat. The technology, which will be put to use through a bot, will search information repositories and give the contextual information quickly, said Vishnu Ramesh, Chief Executive Officer and Co-Founder of Subtl.ai.

Subtl.ai lets its customers build AI search engines automatically out of documents of their choice. “Our machine learning and artificial intelligence technology can take in any question users would normally ask an expert. The engine will then retrieve information most contextually relevant to them from all of their data. We now have one big State-owned bank as our paying client. We are now pitching it for educational institutions. We are focused on improving the relationship the enterprises have with their data,” he said.

---

The underlying technology was built based on research experience in areas of natural language processing and question and answer systems at IIIT-H. The bot can give relevant answers even if questions are reframed as it is based on ‘semantics understanding’. The bot learns with every query- it makes a relationship and allows the user to jump between documents.

Subtl.ai combines various deep tech, document comprehension models, cognitive modelling, relevance engine, natural language processing solution, question and answer systems to help the users find what they intend to find in a document. It will find use in auditing, due diligence, research, and newsrooms. Bots can be customised to new verticals, he said adding that the company is currently solving problems for banks in India and the ed-tech sector.

“Bankers in India have a lot of documents that contain relevant information such as the latest RBI regulations and banking process manuals. We are pioneering this use case in India for the first time with the banking client. We have started sales process for other banks in India and will be scaling up in this space over 2021,” he said. “In Ed-tech scenario, we can help students find information immediately from textbooks, class notes and websites all at once. We are exploring an engagement with a big edtech player. We are expanding our market presence in India and will enter the US markets in 2022,” Ramesh told ‘Telangana Today’.

The company got rejected by YCombinator, an American startup accelerator that provides seed funding (earliest funding), twice. “The first time we got rejected, we realised that we should have more market focus to increase our chances of success. And the second time, we realised we need not run sales funnels only in one domain. Rejection is the best motivator if you are willing to learn from it,” said Ramesh posted on Linkedin recently.

---

Now you can get handpicked stories from Telangana Today on Telegram everyday. Click the link to subscribe.

Click to follow Telangana Today Facebook page and Twitter .

---