Wednesday, May 12, 2027 · 9:00 AM – 9:45 AMMain Stage
Every engineering organisation is really a machine for shortening the distance between a question and its answer. This keynote looks at what changes structurally when that loop drops under ten minutes…
Wednesday, May 12, 2027 · 10:00 AM – 10:30 AMRoom 2A
Our monorepo CI took 40 minutes on a good day. This talk walks through how we cut it to 6 minutes with content-addressed caching, remote execution, and a test-selection model — including the two migra…
PRPriya Raman — Principal Engineer, Latticework Systems
Wednesday, May 12, 2027 · 10:00 AM – 10:30 AMRoom 2B
Canaries, feature flags and automated rollback sound great until the first incident where all three disagree. Here is the decision tree we use, the four metrics that actually gate a rollout, and the p…
HMHugo Martins — Senior Site Reliability Engineer, Vantage Point
Wednesday, May 12, 2027 · 11:00 AM – 1:00 PMWorkshop Lab
A hands-on session where we build a working internal developer platform from an empty repository: golden paths, a service catalogue, a paved-road CI template and a self-service environment. Bring a la…
Three engineering leaders argue about where developer productivity should live: in a platform team, in every team, or nowhere at all. Expect disagreement about metrics and at least one strong opinion…
ABAna Beltrán — VP of Engineering, Northwind Data
KTKeiko Tanaka — Engineering Manager, Lumen Works
FHFarid Haddad — Principal Architect, Meridian Systems
Thursday, May 13, 2027 · 9:30 AM – 10:00 AMMain Stage
Agent evaluations are flaky, expensive and easy to fool. This talk covers the harness we run on every pull request: deterministic seeds, judge calibration, cost budgets, and the tripwires that catch a…
IOIris Oyelaran — Staff Machine Learning Engineer, Corvid AI
Thursday, May 13, 2027 · 11:00 AM – 11:30 AMRoom 2B
Recall@10 on a static corpus tells you almost nothing about a retrieval system that ingests two million documents a day. We cover index rebuild strategies, hybrid search tuning and the operational cos…
Prompts are code with none of the safety. This session shows how we version prompts, snapshot their outputs, score drift with a cheap judge, and gate deploys on the result — plus what happened the wee…
Friday, May 14, 2027 · 10:00 AM – 10:30 AMMain Stage
If your platform has users, it has churn, onboarding and a roadmap. We ran our platform group as a product team for two years: here is the research cadence, the adoption metric we settled on, and the…
Traces for agent runs look nothing like traces for request/response services: they branch, retry and argue with themselves. This talk covers the span model we settled on, cost attribution per step, an…
FHFarid Haddad — Principal Architect, Meridian Systems