This event requires an RSVP; to attend, submit form hereHow can we make small language models efficient enough for low-power devices without sacrificing reliability and task performance? This roundtable will explore why optimizing SLMs requires more than simply choosing a smaller model it requires careful trade-offs across model architecture, compression techniques, hardware constraints, and deployment pipelines. The discussion will cover techniques such as advanced quantization, pruning, sparsity, distillation, and hardware-specific compilation, with an emphasis on what actually translates into latency, memory, and energy-efficiency gains in real-world settings. Participants will discuss how to evaluate quality degradation after optimization, when local inference is appropriate, and how hardware-aware design can improve production readiness. Join this interactive session to share practical lessons and learn how teams are making SLMs faster, lighter, and more deployable in constrained environments.
