We’re excited to announce the winners of Databricks’ inaugural Asia-Pacific Massive Language Mannequin (LLM) Cup, a first-of-its-kind competitors within the area, which garnered participation from over 1,000 knowledge and AI practitioners throughout greater than 10 nations.
From October to December 2023, contributors had been invited to construct LLM-powered functions utilizing Databricks to resolve real-world enterprise issues. To assist arrange contributors for fulfillment, previous to the hackathon, we offered self-paced studying workshops on constructing LLMs and devoted coaching led by Databricks’ answer architects.
Entries had been evaluated primarily based on creativity, enterprise applicability, relevance, thoroughness and the scalability of the LLM structure. Whereas there have been a lot of impactful initiatives, two groups stood out with their modern concepts addressing important organisational challenges in telecommunications and cybersecurity.
Congratulations to the successful groups whose improvements show simply how rapidly groups can rework companies with the proper expertise!
Optus unlocks worth from name centre buyer conversations with LLM, providing 360-degree view of its prospects
Securing the grand prize, Contact Catalyst from Optus, an Australian telecommunications firm, addressed a prevailing enterprise problem: understanding prospects’ wants throughout calls. Name centre brokers sometimes take handbook notes, resulting in unreliable interpretations, typically leading to buyer frustration and dissatisfaction.
Utilizing artificial knowledge units, the group analysed knowledge in an automatic and safe vogue, and navigated the complexities of managing huge volumes of textual content and voice knowledge from buyer interactions and gained deep, real-time insights into prospects’ wants, considerably enhancing the general consumer expertise.
Its LLM can effectively classify the intent of the shopper name, present real-time detection of latest matters, content material summarisation sentiment evaluation and detect aggression, alerting the enterprise to handle the shopper’s issues rapidly and politely.
Ensign InfoSecurity uncovers LLM vulnerabilities to fortify cybersecurity past typical measures
The runner-up, Pink Teaming Your LLM by Ensign InfoSecurity (“Ensign”), Asia’s largest pure-play end-to-end cybersecurity service supplier, demonstrated their answer to counter the cyber vulnerabilities with rising deployment of LLM-based functions.
With larger accessibility to LLM APIs, extra corporations are constructing their very own LLM-based functions. But, the fast deployment of those fashions has outpaced the event of complete safety protocols, leaving many functions weak to high-risks cyberattacks. The Ensign group devised a “crimson teaming method” to coach LLMs with proprietary jailbreak prompts. Consequently, the LLMs proactively recognized and addressed vulnerabilities in LLM defences, enhancing total cybersecurity.
We’re thrilled to see a lot curiosity and innovation coming from the information and AI neighborhood throughout APJ. It is inspiring to see groups use Databricks to develop LLMs that sort out their distinctive challenges – actually attending to the center of what they should higher serve their prospects and revolutionise their enterprise operations.
Each groups offered their successful entries at Databricks’ APJ Information + AI World Tour on Thursday, 18 January 2024 (recording obtainable on-demand).
Go to the venture gallery to be taught extra about their submissions: https://databricks-llm-cup-2023.devpost.com/project-gallery.
