QuBD
NewImproved Kolmogorov/Chaitin/Solomonoff complexity estimator, in theory and in practice.
Pedram Bakhtiarifard
PhD Fellow @ UCPH, Department of Computer Science
bakhpedram@gmail.com / pba@di.ku.dkCopenhagen, Denmark
I’m a PhD Fellow part of the SAINTS Lab led by Prof. Raghavendra Selvan. I work on DL optimization through the lens of algorithmic information theory, optimization geometry, and quantization.
I led the development of CarbonTracker, recognized by the European Commission’s Innovation Radar, and as a TA I taught Programming and Problem Solving and Software Development courses.
Prior to academia, I was a Software Engineer at Danske Bank, where I also built C#/.NET mortgage-refinancing services. Outside the lab, I discuss AI infrastructure and energy use with industry and media.
Projects
MoMos
NewAlgorithmic simplification of neural networks.
CARAML
A framework for aligning AI development with present-day needs and future-facing constraints.
EC-NAS-Bench
NAS-Bench-101 with modeled energy use, making accuracy-energy tradeoffs visible in architecture search.
CarbonTracker
Package for tracking ML training energy and local CO2eq/kWh.
Moments in time
Selected public breadcrumbs: talks, papers, press, and a few milestones.