News
Recent publications, awards, funding, and group updates.
BatchGen accepted to OSDI 2026
BatchGen targets throughput-first inference for very large MoE-style models.
ContextPilot accepted to MLSys 2026
ContextPilot speeds long-context inference through context reuse.
BitDecoding accepted to HPCA 2026
BitDecoding accelerates low-bit KV-cache inference by unlocking Tensor Cores.
MoE-CAP accepted to NeurIPS 2025
MoE-CAP appears in the Dataset and Benchmark Track for mixture-of-experts evaluation.
Award for AI4Math systems research
New funding to build systems that support AI-driven mathematical discovery.
WaferLLM accepted to OSDI 2025
WaferLLM is a wafer-scale LLM inference system designed beyond GPU assumptions.
Promoted to Reader at Edinburgh
Promotion to Reader (Associate Professor) starts in August 2025.
ARIA award for scaling AI compute
Imperial, Edinburgh, and Cambridge team up on modular AI systems simulation.
Tenplex accepted to SOSP 2024
Tenplex enables elastic LLM training with multi-dimensional parallelism.
Yao Fu wins Rising Star in ML and Systems
Yao was selected from a global cohort for systems-focused ML research.
ServerlessLLM accepted to OSDI 2024
A checkpoint-aware serverless inference system for low-latency LLM deployment.
Microsoft Research StarTrack Scholar Award
Received MSR recognition for early-career computer systems research.
ML Systems CDT funded by EPSRC and industry
Edinburgh launched a doctoral training centre spanning the AI systems stack.
TorchOpt accepted to JMLR
TorchOpt also became a PyTorch Ecosystem Project after two years of incubation.
Chancellor's Rising Star finalist at Edinburgh
Recognized through school and college-level selection for rising research leadership.
GEAR accepted to ICML 2023
GEAR bridges reinforcement learning and LLM training workflows.
Ekko accepted to OSDI 2022
Ekko unifies training and inference for low-latency recommender model updates.
MegBA accepted to ECCV 2022
MegBA is a GPU-based distributed library for large-scale bundle adjustment.
Cameo accepted to NSDI 2021
Cameo enables fine-grained real-time stream processing with deadline awareness.
KungFu accepted to OSDI 2020
KungFu makes distributed ML training adaptive and performance-portable.
Invited SOSP AI Systems workshop talk
Talked on adaptive distributed training for deep learning models.