Glossary

Electro-optical transceiver — converts electronic data into an optical signal and vice versa.

Expert parallelism — distributing expert networks of AI models across multiple computing devices to optimize memory usage and processing efficiency.

A laser module that is separate from, and provides optical power to, silicon photonics chips in co-packaged optics modules.

Fat-tree design (fat-tree topology) — a network topology that optimizes communication efficiency.

Figure of merit — a quantitative measure used to evaluate and compare the performance or efficiency of devices, systems, or methods.

Fine-Tuning (AI/ML) — The process of adapting a pre-trained AI model to a specific task or domain by training it on additional data.

Large-scale machine learning models trained on diverse data that serve as a foundation for developing AI applications via transfer learning.

Generative AI (GenAI) — AI technology that creates content from learned patterns.