Vail Systems, Inc. Innovation & Technology Culture

How is your team integrating AI and ML into the product development process, and what specific improvements have you seen as a result?

One example is that we’re developing ML pipelines to assist in filling out requests for proposals. Our system retrieves information from internal knowledge bases and a LLM is configured to present it in the requested format. When equipped with knowledge base results as context, the LLM can determine how our company’s capabilities align with requirements for potential projects. As a result, initial responses to many questions on these forms can be pre-filled before undergoing human review. 

This initiative increases the efficiency with which employees can answer questions about company capabilities. By indexing large sets of documents, the system also provides the opportunity to incorporate relevant information that may have otherwise escaped notice.

What strategies are you employing to ensure that your systems and processes keep up with the rapid advancements in AI and ML?

We seek out new publications via literature review, attend academic conferences and industrial meetups, subscribe to AI and ML mailing lists and keep an eye on the latest projects from different companies and leaderboard sites, such as Hugging Face. The next step is thinking about how we could apply these findings to our own use cases. Our research group also hires PhD interns with diverse academic backgrounds from different universities, and we publish our results whenever possible. 

Ultimately, the key aspect that keeps this system running is our team’s collaborative culture. We’re always sharing the things we discover, whether it’s in an email thread or an impromptu discussion over lunch. These exchanges often naturally spiral into brainstorming sessions that can last several days.

Can you share some examples of how AI and ML has directly contributed to enhancing your product line or accelerating time to market?

Our offerings to clients incorporate tools that extract and analyze the subjects of conversations that take place on our platform. New advances in ML have improved the quality of automated summaries and unlocked the possibility of easily answering specific questions about the data at scale. Our team is also developing solutions that can synthesize multiple sources of information in real time to provide the most helpful response during a phone call.

Some enhancements to our product line leverage tools that we’ve built in anticipation of future requirements. We’ve developed a robust speaker recognition system, which can validate the identity of a caller over the phone. This will be useful for some of our clients that want to implement additional security measures. Moreover, due to the increased risk of fraudulent callers and voice spoofing, we’ve created novel models to identify fake voices in telephonic environments. Typical customers are still in the early stages of considering this problem, but as soon as their need arises, we’ll have solutions to offer them.