Research Scientist - Voice AI Foundations
Deepgram- Full Time
- Mid-level (3 to 4 years), Senior (5 to 8 years)
Candidates should have extremely strong technical experience in training Generative AI models. A demonstrated engineering ability in Python and Pytorch is required, along with experience building ML models from scratch in Pytorch. Academic or professional experience with generative multimodal models such as Diffusion Models and GANs, as well as deep learning concepts such as Transformers, is essential. Familiarity with Linux clusters, systems, and scripting is a plus, as is experience working with large distributed systems greater than 100 GPUs.
The Senior Research Engineer will lead and contribute to cutting-edge research in multimodal foundation models. They will design, develop, and experiment with novel algorithms, architectures, and techniques to enhance the performance, efficiency, and scalability of AI models. Additionally, they will optimize model performance for deployment in production environments, focusing on latency, throughput, and computational efficiency without compromising accuracy or robustness. The role also involves inspecting and managing large-scale data clusters to identify inefficiencies and bottlenecks in model training and data loading, as well as collaborating with cross-functional teams including data, applied research, and infrastructure.
Develops multimodal AI technologies for creativity
Luma AI develops multimodal artificial intelligence technologies that enhance human creativity and capabilities. Their main product, the Dream Machine, allows users to interact with various types of data, enabling creative professionals, businesses, and developers to explore innovative applications of AI. Unlike many competitors, Luma AI focuses on integrating multiple modes of interaction, which broadens the possibilities for users. The company operates on a subscription model, providing access to its AI tools and services, and aims to lead the way in AI-driven creativity and productivity.