Irvine Cloud & AI Architecture Reading Group

8 people · led by Arjun Mehta · Tuesdays at 6:30pm · Irvine, CA

Software architects, SREs, and ML engineers in Orange County reading primary systems papers and AWS/Kubernetes production case studies in person.

Group note: One foundational distributed systems or ML infrastructure paper per week. 15-minute presenter overview, 45-minute architecture whiteboard session.

Next for this groupTuesday, November 3

PagedAttention & High-Throughput LLM Serving Architecture

Read the vLLM PagedAttention paper (Sections 2–5) on virtual memory paging for KV cache blocks.

Questions to review

How does block-level KV paging eliminate internal and external memory fragmentation across variable-length requests?

Recent & Upcoming Study Arc

3 sessions

Dynamo & Consistent Hashing — Partitioning, Sloppy Quorums & Vector Clocks
CompletedTuesday, October 20 at 6:30pm
Read Sections 4–6 of the Dynamo paper. Compare tunable R + W > N quorums against modern multi-region cloud tables.
Distributed Consensus & Raft Paper Walkthrough
CompletedTuesday, October 27 at 6:30pm
Read Ongaro & Ousterhout, "In Search of an Understandable Consensus Algorithm (Extended Version)", Sections 1–6 (Leader Election, Log Replication, Safety).
PagedAttention & High-Throughput LLM Serving Architecture
Up nextTuesday, November 3 at 6:30pm
Read the vLLM PagedAttention paper (Sections 2–5) on virtual memory paging for KV cache blocks.

Pull up a chair

Open seats — joining brings you straight to the group’s plan, calendar, and food sign-up.

Create an account and join I already have an account

Who’s in the group

8 people

Group host & leader

Arjun Mehta

Distributed systems engineer in Irvine Spectrum. We read one systems or AI architecture paper each week and whiteboard the trade-offs.

Members (first names):

AMArjunAPAudreyBHBenjaminDCDavidEBEthanNRNatalieOFOwenPPPriya

Shared roles & hospitality

HostingPresentingQuizmasterSnackDrinks