Conversational messaging app, Gupshup, has launched the ACE LLM family of domain-specific Large Language Models (LLM). It is specialised for functions like as marketing, commerce, support, HR and IT, and industries including banking, retail and utilities.

Built upon foundation models such as Meta’s Llama 2, OpenAI GPT-3.5 Turbo, Mosaic MPT, Flan T-5, and others, it can be used for specific industries and functions. It is also equipped with enterprise-grade safety controls and guardrails. 

The LLM facilitates enterprises to quickly and effectively transform conversational experiences across various stages of the customer lifecycle. From product discovery, lead generation, and commerce to troubleshooting and customer support, it enables more precise, human-like interactions delivered with speed, scale and compliance. This encompasses data governance while maintaining a lower total cost of ownership. 

Available in 7 to 70 billion parameter sizes, ACE LLM generates text in 100 plus languages such as Spanish, Portuguese, French, German, Bahasa, Arabic, Mandarin, Hindi, and English, among other languages.

The built-in guardrails in ACE LLM eliminate irrelevant, out-of-context responses. Additionally, it supports enterprise-level controls for accuracy, source data boundaries, tone, auditing, teach mode for non-generative responses, automated testing, and analytics.   

Beerud Sheth, CEO and co-founder of Gupshup, said, “To harness the full power of foundation LLMs, enterprises need to fine-tune them for their domain requirements and add additional guardrails around security, compliance, and relevance, while also ensuring data residency and cost efficiency.”