Key Takeaways
Since it was founded in 2018, the Scottish FinTech startup Aveni has made a name for itself, developing AI solutions for the financial services industry. But so far, the Scottish FinTech startup has limited its efforts to modifying existing models such OpenAI’s GPT range and the open-source RoBERTa.
After closing an £11 million Series A, however, Aveni is preparing to take things up a notch with plans to build a new large language model (LLM) from scratch.
In a statement announcing the latest funding round, an Aveni representative said:
“This investment will enable Aveni to build on the success of existing products, further establishing its presence across the financial services sector and introducing revolutionary technology through the creation of FinLLM, a financial services specific large language model.”
The new “industry-aligned” model is being developed in partnership with Aveni investors Lloyds Banking Group and Nationwide, which rank among the UK’s largest financial institutions.
By building an LLM from the ground up, the initiative could overcome some challenges that have limited banks’ adoption of generative AI until now.
From customer service chatbots to fraud detection systems, AI developers like Aveni have applied LLMs to a variety of banking applications.
But as Arun Subramaniyan, CEO of AI startup Articul8, explained to CCN:
“General-purpose models are necessary, but not sufficient […] Today, there’s no reason to go build yet another model that just understands language, that’s done. However, it’s one thing to understand language, you still need to understand the domain.”
Going forward, Subramaniyan said banks will need more tailored solutions like the one planned by Aveni to meet their specific security and compliance needs.
In the era of general-purpose AI, Google, OpenAI, and a handful of other large companies have done the bulk of the legwork in developing new models. But as industry-specific AI gathers steam, smaller startups like Aveni could shine.
Besides building new LLMs from the ground, another important part of the conversation is what Aveni has been doing up until now—melding together existing models to meet sector-specific needs.
Subramaniyan predicted that a mixed approach will prevail, with banks incorporating a variety of different models into dynamic AI architectures.
Ultimately, the future of finance could see FinLLM working alongside ChatGPT in a multi-layered technology stack that differs depending on the specific use case.