A true conversational AI system engages in human-like dialogue, understands context, and provides intelligent responses. Such AI models are often huge and highly complex. With NVIDIA GPUs and NetApp storage, massive, state-of-the-art language models can be trained and optimized to run inference rapidly. This is a major stride towards ending the trade- off between an AI model that is fast versus one that is large and complex. GPU-optimized language understanding models can be integrated into AI applications for industries such as healthcare, retail, and financial services, powering advanced digital voice assistants in smart speakers and customer service lines. These high-quality conversational AI systems allow businesses across verticals to provide previously unattainable personalized services when engaging with customers.
Jarvis enables the deployment of use cases such as virtual assistants, digital avatars, multimodal sensor fusion (CV fused with ASR/NLP/TTS), or any ASR/NLP/TTS/CV stand-alone use case, such as transcription. We built a virtual retail assistant that can answer questions regarding weather, points-of-interest, and inventory pricing. We also demonstrated how to improve the natural language understanding capabilities of the conversational AI system by archiving conversation history using Cloud Sync and training NeMo models on new data.