PerfDojo: An AI coach for faster ML on any chip

Kari Jaaskelainen

06 Nov 2025 — 1 min read

In a nutshell

Making ML run fast on every chip is hard. Different CPUs, GPUs, and accelerators need different tricks, especially with features like sparsity and quantization. Manual tuning is slow; many automatic tools are opaque.

What’s new

PerfLLM + PerfDojo turn optimization into a reinforcement learning game guided by large language models.
They use a human-readable, math-inspired code form; transformations guarantee semantic validity while exploring faster variants.
Works without prior hardware-specific knowledge, enabling both human insight and effective agent training.

Why it matters

Portable speed-ups: the paper reports gains across CPUs (x86, Arm, RISC-V) and GPUs.
Fewer black-box heuristics; more interpretable optimization.

Automated performance without arcane hardware hacks.

By Andrei Ivanov, Siyuan Shen, Gioele Gottardo, Marcin Chrapek, Afif Boudaoud, Timo Schneider, Luca Benini, Torsten Hoefler. Read more: http://arxiv.org/abs/2511.03586v1

Paper: http://arxiv.org/abs/2511.03586v1

Register: https://www.AiFeta.com

#MachineLearning #AI #Performance #LLM #ReinforcementLearning #CPUs #GPUs #RISCV #Arm #Systems #HPC #Research

Automating GDPR Compliance: A Roadmap for Companies and Law Firms

GDPR compliance is more than checkboxes. A new roadmap from the Privatech project shows how automation and machine learning can help companies and law firms assess—and even generate—privacy compliance. * Shift the focus to data processors’ real workflows: drafting policies, mapping data uses, documenting decisions. * Break compliance into machine-ready

FPGAs for Faster, Leaner Deep Learning: A Review of CNN Accelerators

Deep learning drives image search, robots, and medical scans. Most systems lean on CPUs and GPUs. This review asks: what if we run convolutional neural networks (CNNs) on FPGAs—reconfigurable chips you can tailor to the model? * Why FPGAs: custom dataflows, low latency, and strong energy efficiency—great for cameras,

Dynamic-K: Recommendations That Know When to Stop

Most apps show a fixed number of “top” items—say 10 movies or 20 products—assuming there are always enough good options. But that’s not always true: sometimes there are few relevant items, or some users are extra picky. The result? Filler recommendations. Dynamic-K flips the script. Instead of

Teaching chatbots to stop contradicting themselves (DECODE)

Teaching chatbots to stop contradicting themselves Ever had a bot say one thing, then the opposite a few turns later? This study introduces DECODE—a new task and dataset for spotting contradictions in everyday conversations, drawn from both human-human and human-bot chats. * New data beats existing natural language inference (NLI)