Coursera Spring Sale
40% Off Coursera Plus Annual!
Grab it
Explore the journey of building an AI-powered translation company that successfully competes with tech giants in this 43-minute podcast episode featuring DeepL's CEO and founder Jarek Kutylowski. Discover how DeepL launched neural machine translation in 2017, built custom GPU data centers, and developed strategies for small teams to challenge established players like Google Translate. Learn about the technical challenges of AI translation, including why high-quality translations still require human context and how models are tailored for enterprise applications. Examine DeepL's approach to training on curated multilingual datasets, their methods for measuring translation quality, and their stance on avoiding customer-specific fine-tuning. Gain insights into the evolution of speech translation technology, the growing demand for translation services, and the return on investment of incremental quality improvements. Understand how DeepL handles hallucinations in translation models, supports smaller languages and language pairs, and manages large-scale inference infrastructure. Delve into the future role of human translators, the broader impact of AI on global communication, and how translation companies adapt to the evolving large language model landscape while meeting enterprise needs.
