Amir Yazdanbakhsh
RESEARCH SCIENTIST · GOOGLE DEEPMIND
Teaching machines to build better machines.
Embedded Files
I’m a research scientist at Google DeepMind, where I co-founded and co-lead the Machine Learning for Computer Architecture team. I build AI agents that write, optimize, and verify performance-critical software and hardware.
My recent work focuses on agentic code optimization. ECO, our LLM-driven optimizer deployed across Google’s production fleet, has landed more than 6,400 commits and saved the equivalent of several hundred thousand CPU cores. I also use AlphaEvolve to discover compiler heuristics, warehouse-scale code layouts, and FHE kernels for TPUs. With multi-agent systems, I helped build a cycle-accurate RISC-V CPU simulator from scratch that boots an operating system.
Earlier, I co-authored Self-Refine, one of the first works to show that LLMs can improve their own outputs through self-feedback. It was an early step toward recursive self-improvement and has been cited more than 7,000 times.
My work has earned multiple IEEE Micro Top Picks honors, and I’m an inductee of the ISCA Hall of Fame and a Senior Member of IEEE and ACM. I received my Ph.D. from Georgia Tech, where my research was supported by the Microsoft Research PhD Fellowship and the Qualcomm Innovation Fellowship.
I host student researchers. Reach out with a short note on your interests.
RECENT
NEW Aug 2026 · Multi-agent RISC-V simulator boots an OS
Jul 2026 · ECO at OSDI 2026
Jul 2026 · QuArch and ParEVO at ICML 2026
Publications
Publications
Selected work below. The complete list is on Google Scholar.
SELECTED
ECO: An LLM-Driven Efficient Code Optimizer for Warehouse-Scale Computers
OSDI 2026 · pdf · Deployed across Google's fleet
Self-Refine: Iterative Refinement with Self-Feedback
NeurIPS 2023 · pdf · site · 7,000+ citations
Learning Performance-Improving Code Edits
ICLR 2024 · pdf · site · Spotlight · IEEE Micro Top Picks
Concorde: Fast and Accurate CPU Performance Modeling with Compositional Analytical-ML Fusion
ISCA 2025 · pdf · IEEE Micro Top Picks
LIA: A Single-GPU LLM Inference Acceleration with Cooperative AMX-Enabled CPU-GPU Computation and CXL Offloading
ISCA 2025 · pdf · code · IEEE Computer Society Best Paper Award
MESA: Microarchitecture Extensions for Spatial Architecture Generation
ISCA 2023 · pdf · ISCA Hall of Fame
AGENTS FOR CODE & SYSTEM OPTIMIZATION
SWE-fficiency: Can Language Models Optimize Real-World Repositories on Real Workloads?
ICML 2026 · pdf · site
ParEVO: Synthesizing Code for Irregular Data: High-Performance Parallelism through Agentic Evolution
ICML 2026 · pdf · code
AI-Propeller: Warehouse-Scale Interprocedural Code Layout Optimization with AlphaEvolve
CAIS 2026 Workshop · pdf
Adapting AlphaEvolve to Optimize Fully Homomorphic Encryption on TPUs
MLArchSys @ISCA 2026 · pdf
Magellan: Autonomous Discovery of Novel Compiler Optimization Heuristics with AlphaEvolve
C4ML Workshop @HPCA 2026 · pdf
DSTBench: Distributed Systems Traces-Driven Agentic Performance Debugging in Microservices
IC2E 2026
CodeRosetta: Pushing the Boundaries of Unsupervised Code Translation for Parallel Programming
NeurIPS 2024 · pdf · site
EFFICIENT LLMs
OPTIMA: Optimal One-Shot Pruning for LLMs via Quadratic Programming Reconstruction
ACML 2026 · pdf
Learning to Keep a Promise: Scaling Language Model Decoding Parallelism with Learned Asynchronous Decoding
ICML 2025 · pdf · thread
SLIM: One-Shot Quantization and Sparsity with Low-Rank Approximation for LLM Weight Compression
ICML 2025 · pdf
SparseLoRA: Accelerating LLM Fine-Tuning with Contextual Sparsity
ICML 2025 · pdf · thread
Effective Interplay between Sparsity and Quantization: From Theory to Practice
ICLR 2025 · pdf · Spotlight
The Journey Matters: Average Parameter Count over Pre-training Unifies Sparse and Dense Scaling Laws
ICLR 2025 · pdf · thread
SLOPE: Double-Pruned Sparse Plus Lazy Low-Rank Adapter Pretraining of LLMs
ICLR 2025 · pdf
Progressive Gradient Flow for Robust N:M Sparsity Training in Transformers
CPAL 2025 · pdf · code · Oral
ML FOR COMPUTER ARCHITECTURE
QuArch: A Benchmark for Evaluating LLM Reasoning in Computer Architecture
ICML 2026 · pdf · site · Best of CAL at HPCA 2026
Architecture Gym for Benchmarking Machine-Learning Aided Design
ISCA 2023 · pdf · code · video
GRANITE: A Graph Neural Network Model for Basic Block Throughput Estimation
IISWC 2022 · pdf · code
An Evaluation of Edge TPU Accelerators for Convolutional Neural Networks
IISWC 2022 · pdf
HARDWARE FOR AI
RAGO: Systematic Performance Optimization for Retrieval-Augmented Generation Serving
ISCA 2025 · pdf · code
DaCapo: Accelerating Continuous Learning in Autonomous Systems for Video Analytics
ISCA 2024 · pdf · Distinguished Artifact Award
Tandem Processor: Grappling with Emerging Operators in Neural Networks
ASPLOS 2024 · pdf · code · IEEE Micro Top Picks Honorable Mention
In-Storage Domain-Specific Acceleration for Serverless Computing
ASPLOS 2024 · pdf · code
Sparse Attention Acceleration with Synergistic In-Memory Pruning and On-Chip Recomputation
MICRO 2022 · pdf
Accelerating Attention through Gradient-Based Learned Runtime Pruning
ISCA 2022 · pdf
SnaPEA: Predictive Early Activation for Reducing Computation in Deep Convolutional Neural Networks
ISCA 2018 · pdf
General-Purpose Code Acceleration with Limited-Precision Analog Computation
ISCA 2014 · pdf · retrospective · IEEE Micro Top Picks Honorable Mention
LLM REASONING
What Makes Chain-of-Thought Prompting Effective? A Counterfactual Study
EMNLP 2023 · pdf · ArXiv · code
Honors
Honors
IEEE Micro Top Picks · 2025, 2026
IEEE Micro Top Picks Selection Committee · 2026, 2027
IEEE Computer Society Best Paper Award · 2024
ISCA Hall of Fame · 2023
Senior Member, IEEE and ACM
Gold Medal, ACM Student Research Competition · 2018
Microsoft Research PhD Fellowship · 2016–2018
Qualcomm Innovation Fellowship · 2014–2015
IEEE Micro Top Picks Honorable Mention · 2016, 2026
Media Coverage
Media Coverage
Need to Fit Billions of Transistors on a Chip? Let AI Do It, WIRED · 2021
Google's Deep Learning Finds a Critical Path in AI Chips, ZDNet · 2021
Google's Apollo AI for Chip Design Improves Deep Learning Performance by 25%, InfoQ · 2021
Google's AI Lab Unveiled A New Framework for Efficient Chip Design, Analytics Insight · 2021
Writing & Talks
Writing & Talks
Teamwork: When AI Becomes a Research Partner
Google Antigravity Blog · 2026 · read
An Open-Source Gymnasium for ML-Assisted Computer Architecture Design
Google Research Blog · 2023 · read
Architecture 2.0: Data-Centric AI Gymnasium for Computer Architecture Design
OCP Global Summit · 2023 · video
Offline Optimization for Architecting Hardware Accelerators
Google Research Blog · 2022 · read
Machine Learning for Computer Architecture
Google Research Blog · 2021 · read
Massively Large-Scale Distributed Reinforcement Learning with Menger
Google Research Blog · 2020 · read
Page updated
Google Sites
Report abuse