Berkeley, California
UC Berkeley
Berkeley builds the plumbing. Most of what runs underneath the products in this tracker — the serving layer, the distributed runtime, the instruction set, the open evaluation — started in a Berkeley lab and left as open source.
Key findings
- 2023
vLLM and PagedAttention
Kwon, Li, Zhuang, Sheng, Stoica et al., Sky Lab
Applied virtual-memory paging to the KV cache, eliminating the fragmentation that wasted most GPU memory during LLM serving.
Why it matters here · The default open serving engine. When a product in the catalog quotes tokens-per-second or self-hosts a model, this is usually the layer doing it.
2–4× throughput at same latencyvLLM → - 2023
Vicuna and Chatbot Arena
LMSYS / Sky Lab
An open fine-tune plus a blind, crowd-sourced pairwise ranking that turned model comparison into an Elo leaderboard.
Why it matters here · The most-cited independent counterweight to vendor benchmark claims; a model's Arena placement now moves procurement decisions.
- 2017
Ray
Moritz, Nishihara, Stoica et al., RISELab
A distributed runtime for actor-style Python workloads: training, tuning, serving and RL on one substrate.
Why it matters here · Used to train and post-train frontier models, including RLHF pipelines at several labs in this catalog.
- 2010
Apache Spark
Matei Zaharia et al., AMPLab
In-memory resilient distributed datasets, an order of magnitude faster than disk-based MapReduce.
Why it matters here · Still the data-preparation layer under a large share of enterprise AI pipelines.
- 2010
RISC-V
Asanović, Patterson, Waterman, Lee
A free, open, extensible instruction set architecture with no licensing regime.
Why it matters here · The escape hatch from x86/Arm licensing — increasingly the control core inside AI accelerators and robotics controllers.
- 1980 · 1988
Berkeley RISC and RAID
David Patterson, Randy Katz, Garth Gibson
Reduced instruction sets, and redundant arrays of inexpensive disks — commodity parts arranged to beat expensive monoliths.
Why it matters here · The architectural argument the entire hyperscale build-out rests on: scale cheap parts rather than buy big ones.
- 2015
TRPO / GAE, then visuomotor policies
Schulman, Levine, Abbeel, Jordan
Stable policy-gradient methods, and the first end-to-end training of raw pixels to robot torques.
Why it matters here · TRPO's successor PPO is the algorithm behind RLHF; the visuomotor line is the direct ancestor of today's robot foundation models.
- 1985 · Nobel 2025
Macroscopic quantum tunnelling in Josephson junctions
John Clarke, Michel Devoret, John Martinis
Demonstrated that an electrical circuit can behave as a single quantum object with discrete energy levels.
Why it matters here · The experimental basis for every superconducting qubit shipping today — IBM, Google Willow, Rigetti. Awarded the 2025 Nobel Prize in Physics.
- 2014
Caffe
Yangqing Jia, BAIR
The first widely-used fast CNN framework with a model zoo of pretrained weights.
Why it matters here · Established pretrained-weight sharing as normal practice — the cultural precondition for open model hubs.
On the history timeline
Milestones on the 1943 → today timeline credited to this institution.
- 1965Fuzzy SetsLotfi ZadehSymbolic era & the winters
- 1980The Case for the Reduced Instruction Set ComputerDavid Patterson & David DitzelSymbolic era & the winters
- 19834.2BSD ships TCP/IP socketsComputer Systems Research Group (Bill Joy et al.)Symbolic era & the winters
- 1985Macroscopic quantum tunnelling in Josephson junctionsJohn Clarke, Michel Devoret & John MartinisSymbolic era & the winters
- 1988RAID — Redundant Arrays of Inexpensive DisksDavid Patterson, Garth Gibson & Randy KatzConnectionist revival
- 1993Quantum complexity theory and BQPEthan Bernstein & Umesh VaziraniConnectionist revival
- 1995Artificial Intelligence: A Modern ApproachStuart Russell & Peter NorvigConnectionist revival
- 2010Apache Spark and the AMPLab data stackMatei Zaharia, Ion Stoica & the AMPLabDeep learning boom
- 2010RISC-V — an open instruction setKrste Asanović, Andrew Waterman, Yunsup Lee & David PattersonDeep learning boom
- 2014Caffe — the first mainstream deep learning frameworkYangqing Jia & Berkeley AI ResearchDeep learning boom
- 2015Trust region policy optimisation & GAEJohn Schulman, Sergey Levine, Philipp Moritz, Michael Jordan & Pieter AbbeelDeep learning boom
- 2016End-to-end training of visuomotor policiesSergey Levine, Chelsea Finn, Trevor Darrell & Pieter AbbeelDeep learning boom
- 2017Ray — a distributed runtime for AIPhilipp Moritz, Robert Nishihara, Ion Stoica & RISELabTransformer & scaling
- 2023Vicuna and the LMSYS Chatbot ArenaLMSYS / Sky Computing LabGenerative & agentic
- 2023vLLM and PagedAttentionWoosuk Kwon, Zhuohan Li, Ion Stoica et al.Generative & agentic
What to follow
Berkeley AI Research (BAIR)
Research lab · Continuous blog + papers
Robot learning, deep RL, open models and evaluation — the fastest path from Berkeley research to shipped open source.
Sky Computing Lab
Research lab · Continuous
Successor to AMPLab and RISELab. Home of vLLM, SkyPilot and Chatbot Arena infrastructure.
Center for Responsible, Decentralized Intelligence (RDI)
Research lab · Courses, agent benchmarks
LLM agents, agent safety and decentralised systems, including the widely-followed LLM Agents MOOC.
Emerging Technology Management (Haas Exec Ed)
Program · Cohort-based
The management-side counterpart: evaluating, adopting and governing emerging tech inside an organisation.
Berkeley Emerging Technologies Association (BETA)
Community · Semester programming
Student-run projects, speaker sessions and industry work — an early read on what the next cohort is building.