Principal AI Solutions Software Engineer, SambaNova Systems

• Bachelor's or higher degree in Computer Science, Electrical Engineering, Applied Mathematics, Physics, or Statistics
• 5+ years of industry experience, including 3+ in one or more of the following:
• Deep learning algorithm development
• Compiler
• Software-Hardware Co-design
• Proficiency in Python or C++, with a solid foundation in data structures, algorithms, and machine learning
• Proficiency in one or more of the popular ML frameworks (PyTorch/Tensorflow/JAX)
• Experience in machine learning productization and pipeline development in software engineering
• Real-world experience in multi-lingual LLM applications training and inference
• Real-world experience in vision-language multimodality
• Development and deployment Model Training and Inference at scale, Synthetic Data, Information Retrieval, Machine reading comprehension, RLHF/RLAIF, Question Answering, Copilot
• Development with DeepSeed, Megatron, vLLM, and TensorRT

Desirable

• This role offers an unparalleled opportunity to work E2E (end-to-end) with cross-functional teams across Sambanova's software, hardware and ML teams to transform cutting-edge AI research into customer-empowering solutions with superior AI System performance and accuracy, propelling us towards making a substantial impact across diverse industries
• Work with both external users and internal engineering teams to develop robust, efficient and scalable AI solutions that meet customers' needs
• Integrate the latest model architecture, data curation, and performance optimization technologies from the AI industry and research into SambaNova's technical stack
• Create end-to-end solutions for ML applications to enable model training and fine-tuning on domain-specific data
• Enable high throughput and low latency inference applications for at-scale deployment
• Collaborate with cross-functional software and hardware teams to innovate customer-centric applications
• Work with diverse data types: textual, unstructured, tabular and multimodal data