From Small-Town to YC Success, Kastle’s AI Agents Shape the Mortgage Industry

Kastle AI

Y Combinator, or YC, the world’s largest startup accelerator, has accelerated funding for selective promising startups increasing to four batches this year. Interestingly, a number of startups founded by Indians or Indian-origin founders have been part of YC batches. One such startup from the recent YC Summer ‘24 batch is Kastle, founded by Indian-origin entrepreneurs, which aims to bring AI agents to the mortgage sector.

Rishi Choudhary, the co-founder of Kastle, has turned his journey from a small town in Orissa to Silicon Valley into a tale of ambition and innovation in the mortgage industry. With a focus on leveraging AI for customer service in finance, Choudhary along with co-founder Nitish Poddar, started Kastle to bring in AI voice agents to automate and improve customer interactions.

Kastle founders Rishi Choudhary and Nitish Poddar with YC partner Diana Hu. Source: Linkedin

From Real Estate to AI for Mortgage Servicing

“We’re building agents that can handle complex financial tasks and workers, and they all live in this castle, which is super secure,” explained Choudhury in an exclusive interaction with AIM, while discussing the origin of the name.

Choudhary’s path to entrepreneurship began with an internship at Redfin, a tech-powered real estate brokerage in the U.S. There, he discovered the vast potential of the mortgage industry, which holds approximately $14 trillion in outstanding mortgage debt.

“I learned about how old this industry is, and how big and central it is to the global financial system,” said Choudhury. His experience there made him realise the potential of the mortgage sector that was waiting to be disrupted by AI.

After graduating with a degree in systems engineering from the University of Illinois, Choudhary moved to San Francisco, where he reconnected with his co-founder and CTO Poddar, who shared his entrepreneurial spirit.

“We started out building software for real estate agents, but quickly pivoted to focus on mortgage loan officers,” he explained. This shift was a response to the crowded market and the unique challenges faced by mortgage professionals.

He even spoke fondly about his YC experience. “The learning that happens in Y Combinator is immense. You’re surrounded by talented people who are facing similar challenges,” he said. This collaborative environment has been instrumental in shaping Kastle’s direction and growth.

Concoction of LLMs

Kastle utilises large language models (LLMs) to enhance customer interactions in the mortgage sector. The company self-hosts these models, fine-tuning them specifically for mortgage-related queries. “We use a structured workflow system that enables the AI to navigate conversations effectively,” Choudhary noted.

The startup has fine-tuned Llama-8B for its AI model and is using Eleven Labs for text-to-speech.

Their models are trained on extensive datasets, including mortgage-related documents and customer interactions, allowing them to respond accurately to a wide range of inquiries.

The agentic workflow developed by Kastle allows for a high degree of customisation based on client needs. “We have built guardrails to ensure compliance, which significantly reduces the chance of hallucinations,” he added.

Furthermore, Kastle aims to minimise latency in voice interactions, targeting a response time under one second to make conversations feel ‘superhuman.’

Choudhary emphasises the importance of this speed in creating a seamless customer experience. This allows borrowers to receive instant answers to their questions, whether they are inquiring about loan status, interest rates, or documentation requirements.

Adoption of AI in Finance

Leveraging AI and LLMs in the financial sector has seen increasing adoption. For instance, a Bengaluru-based startup, ‘OnFinance’ is using AI to bring solutions for BFSI sectors, where their proprietary LLM will help analysts, advisors and other companies.

Similarly, Gnani AI, an Indian startup is building voice-based small language models in 12 Indic languages. The models will cater to banking, finance, security and insurance domains.

However, Kastle looks to target the mortgage industry alone, a sector that has a huge market in the US.

Choudhury observed the mortgage industry to be traditionally slow to adapt to technological advancements. Many mortgage loan officers still rely on outdated processes that can frustrate customers and lead to inefficiencies. Choudhary recognised this gap and set out to create a solution that would not only streamline operations but also enhance the customer experience.

“Our mission is to empower mortgage professionals by providing them with tools that enhance their efficiency and enable them to focus on building relationships,” Choudhary stated.

From the US To India?

Choudhary’s ambitions extend beyond the U.S. market. Both he and his co-founder, who also hails from India, are keen on bringing their innovative solutions to their home country. “We want to take this company to India because our parents are there. All our families are there,” he remarked.

The potential to expand into insurance, banking, and small business lending in India is part of a broader vision that seeks to adapt Kastle’s technology to different financial contexts.

The post From Small-Town to YC Success, Kastle’s AI Agents Shape the Mortgage Industry appeared first on AIM.

Follow us on Twitter, Facebook
0 0 votes
Article Rating
Subscribe
Notify of
guest
0 comments
Oldest
New Most Voted
Inline Feedbacks
View all comments

Latest stories

You might also like...