Transformed: How to Build a Successful NLP Product with Transformers
Transformers and natural language processing (NLP) are frequently viewed as a match made in heaven. Over the past few years, this type of machine learning models has become a default answer to any language-related task, would that be summarization, machine translation, or text generation. Transformers are a no-brainer. But is it really so? In her talk, Vera will discuss pragmatic considerations for building Transformer-powered NLP products. What makes Transformers a wise business decision? What are the costs involved? How can one build a viable business and product with Transformers? Vera will touch on these and other practical aspects of bringing Transformers from news headlines to real-life production.
Key Takeaways
• When it comes to Product, a pure transformer language model is not a substitute for a value proposition;
• Modern sophisticated transformer models come with a hefty price tag (even the open-source ones);
• Can transformers be an element of a business model? Yes, they should be.
Vera Serdiukova is a Senior AI Product Manager at Salesforce, where she works primarily in the field of Natural Language Processing (NLP). Before Salesforce, she was a part of LG’s Silicon Valley lab, where her focus was on Edge AI/on-device Machine Learning. Prior to that, Vera built Conversational AI interfaces for Bosch’s robotics, connected car, and smart home products.
