Nastaran Darabi
Papers
5
Total Citations
46
H-Index
3
About
Nastaran Darabi is a researcher at the forefront of edge intelligence, uncertainty-aware computing, and autonomous robotics, with a focus on making AI systems more reliable and energy-efficient in real-world deployments. Her work bridges hardware and algorithmic innovation, tackling the fundamental challenge of deploying robust deep learning at the resource-constrained edge. Darabi's most influential contribution, MC-CIM (2022, 32 citations), introduced a compute-in-memory framework that leverages Monte Carlo dropout methods to enable Bayesian deep neural networks capable of expressing prediction uncertainty — a critical capability for safety-sensitive applications. This work elegantly combines hardware efficiency with principled probabilistic inference. Building on this foundation, her STARNet framework addresses sensor trustworthiness in autonomous systems, developing lightweight anomaly recognition for complex sensors like LiDAR and RADAR that are vulnerable to real-world failure modes. Her research on uncertainty-aware pose estimation for insect-scale drones further demonstrates her commitment to practical, miniaturized autonomy. Through her broader program examining the sensing-to-action loop in edge robotics and smart systems, Darabi is shaping how autonomous platforms perceive, reason, and act under uncertainty. With a growing citation record and increasingly ambitious research scope, she represents an emerging voice in dependable edge AI.
Research Focus
Key Achievements
Top Papers
- 1
- 2
- 3
- 4
- 5