We are pleased to invite all TalTech academic staff to the next session in the TalTech AI monthly research seminar series, taking place on Tuesday, 2 December, from 15:00–16:30 in SOC-312.
This month’s speaker is Maksim Jenihhin, Associate Professor of Computing Systems Reliability and Head of the Trustworthy and Efficient Computing Hardware (TECH) research group at TalTech. His talk “Techniques for Reliability in Edge-AI Chips”, will explore how to ensure reliability and energy efficiency in AI-accelerating edge devices.
Talk abstract:
AI applications increasingly rely on a hybrid cloud–edge computing paradigm, keeping complex model training in the cloud while performing fast, efficient inference at the edge. This shift establishes the concept of Edge AI and introduces new reliability requirements for the underlying hardware, especially in safety- and mission-critical systems. The talk discusses techniques for assessing and improving soft-error and lifetime reliability in Deep Learning accelerators, examines the use of approximate computing and quantization, and highlights architectural considerations for ASIC and FPGA-based accelerators.
The talk will be followed by open discussion and networking over coffee and light snacks. The talk will also be streamed via Teams, but we strongly encourage in-person participation to benefit from the discussions.