Uploaded January 2026 | Updated September 2026, 2 weeks ago
One of the tracks at the HiPEAC 2026 conference in Krakow, Poland, aimed to highlight the work that women are doing in high-performance computing (HPC), AI and critical real-time embedded systems. The aim, according to HiPEAC is to:
• Highlight the work and research of women in HPC.
• Attract female talent to the HPC field.
• Create a collaboration and support network for women working and researching in HPC.
• Empower and support women at all stages of their careers in HPC.
One of the papers presented in this session was by Dolly Sapra, Assistant Professor (Edge AI) at the University of Amsterdam, in which she introduced her work on self-adaptive AI models. The premise is that in edge AI systems, most AI models are designed for fixed execution settings, limiting their efficiency and robustness on real devices.
Her paper illustrated how adaptive inference is a way of overcoming these limitations, whereby an existing AI model can be transformed into an adaptive one. Through elastic, modular model designs, different levels of adaptation can be selected at runtime based on available resources, enabling the model to balance accuracy, latency, and energy consumption.
Her talk highlighted how this resource-aware adaptability enables more efficient and sustainable AI deployment directly on edge devices.
EE Times caught up with Dolly Sapra to learn more about her work. She said that adaptivity itself is a new concept as applied to AI models, and that it’s not just about making smaller and smaller models, but making models that can change and adapt according to the environment and need.
You can hear more about this in the video conversation with Dolly here.
------
See also:
🎙 Listen to EE Times On Air podcasts on power, embedded, AI, and more: eetimes.com/podcasts/?utm_source=eetimes_youtube&utm_medium=social&utm_campaign=description
🎥 Don't miss an episode and subscribe to our YouTube channel: youtube.com/@eetimes747
Stay up to date with the latest news in the electronics industry
➡️ EE Times: eetimes.com/?utm_source=eetimes_youtube&utm_medium=social&utm_campaign=description
➡️ Power Electronics News: powerelectronicsnews.com/?utm_source=eetimes_youtube&utm_medium=social&utm_campaign=description
➡️ Embedded: embedded.com/?utm_source=eetimes_youtube&utm_medium=social&utm_campaign=description
➡️ EDN: edn.com/?utm_source=eetimes_youtube&utm_medium=social&utm_campaign=description
📰 Subscribe to our newsletters: aspencore.dragonforms.com/loading.do?omedasite=EventSubscription&pk=youtube
🖥️ Attend free tech webinars: eetimes.com/webinars/?utm_source=eetimes_youtube&utm_medium=social&utm_campaign=description
📈 Download free industry white papers: eetimes.com/techpapers/?utm_source=eetimes_youtube&utm_medium=social&utm_campaign=description
🎤 Check out our virtual events on EE Times: events.eetimes.com/?utm_source=eetimes_youtube&utm_medium=social&utm_campaign=description
🎉 Interested in advertising opportunities? Reach out to sales@aspencore.com
One of the tracks at the HiPEAC 2026 conference in Krakow, Poland, aimed to highlight the work that women are doing in high-performance computing (HPC), AI and critical real-time embedded systems. The aim, according to HiPEAC is to:
• Highlight the work and research of women in HPC.
• Attract female talent to the HPC field.
• Create a collaboration and support network for women working and researching in HPC.
• Empower and support women at all stages of their careers in HPC.
One of the papers presented in this session was by Dolly Sapra, Assistant Professor (Edge AI) at the University of Amsterdam, in which she introduced her work on self-adaptive AI models. The premise is that in edge AI systems, most AI models are designed for fixed execution settings, limiting their efficiency and robustness on real devices.
Her paper illustrated how adaptive inference is a way of overcoming these limitations, whereby an existing AI model can be transformed into an adaptive one. Through elastic, modular model designs, different levels of adaptation can be selected at runtime based on available resources, enabling the model to balance accuracy, latency, and energy consumption.
Her talk highlighted how this resource-aware adaptability enables more efficient and sustainable AI deployment directly on edge devices.
EE Times caught up with Dolly Sapra to learn more about her work. She said that adaptivity itself is a new concept as applied to AI models, and that it’s not just about making smaller and smaller models, but making models that can change and adapt according to the environment and need.
You can hear more about this in the video conversation with Dolly here.
------
See also:
🎙 Listen to EE Times On Air podcasts on power, embedded, AI, and more: eetimes.com/podcasts/?utm_source=eetimes_youtube&utm_medium=social&utm_campaign=description
🎥 Don't miss an episode and subscribe to our YouTube channel: youtube.com/@eetimes747
Stay up to date with the latest news in the electronics industry
➡️ EE Times: eetimes.com/?utm_source=eetimes_youtube&utm_medium=social&utm_campaign=description
➡️ Power Electronics News: powerelectronicsnews.com/?utm_source=eetimes_youtube&utm_medium=social&utm_campaign=description
➡️ Embedded: embedded.com/?utm_source=eetimes_youtube&utm_medium=social&utm_campaign=description
➡️ EDN: edn.com/?utm_source=eetimes_youtube&utm_medium=social&utm_campaign=description
📰 Subscribe to our newsletters: aspencore.dragonforms.com/loading.do?omedasite=EventSubscription&pk=youtube
🖥️ Attend free tech webinars: eetimes.com/webinars/?utm_source=eetimes_youtube&utm_medium=social&utm_campaign=description
📈 Download free industry white papers: eetimes.com/techpapers/?utm_source=eetimes_youtube&utm_medium=social&utm_campaign=description
🎤 Check out our virtual events on EE Times: events.eetimes.com/?utm_source=eetimes_youtube&utm_medium=social&utm_campaign=description
🎉 Interested in advertising opportunities? Reach out to sales@aspencore.com
