Applied AI
Talks from teams running models in production. How they test them, what they cost, where they fail and how the product around them changes once the demo is over and real users arrive.
Track hostZara HaddadHead of AI Product, HaldenApplied AI is for the work that starts after the prototype. Every talk on this track comes from a team that has put a model in front of paying customers and lived with the consequences: the evaluation suite that caught a regression, the feature that was quietly switched off, the pricing change that followed the first real invoice.
It suits engineers, product managers and designers who already have something in production, or are about to, and want specifics rather than predictions. Expect numbers, failure stories and code on slides, and very little about what might happen in five years.
Zara Haddad curated the track by asking every shortlisted speaker the same question: what would you do differently if you started again tomorrow?
Day 1Wednesday 9 June
Applied AI· Talk · Advanced
Two million users, one agent: the failures we did not test for
Kenji Arai, Corvid
10:30–11:10Main Stage
Applied AI· Workshop · Advanced
Build an evaluation suite for an LLM feature in two hours
Elif Demir, Kenji Arai
14:30–16:30Lab36 seats
Applied AI· Talk · Intermediate
Latency is a design material: interfaces for streaming models
Omar Saleh, Lumio
15:20–16:00Main Stage
Applied AI· Keynote
Trust is the product: two years of shipping AI to people who did not ask for it
Zara Haddad, Halden
16:30–17:20Main Stage
Day 2Thursday 10 June
Applied AI· Talk · Advanced
Your evals are lying to you: measuring models on real work
Elif Demir, Tidewater Labs
11:40–12:20Studio
Applied AI· Panel
Who is accountable when the agent gets it wrong?
Elena Petrova, Zara Haddad and 2 more
12:30–13:15Main Stage
Applied AI· Talk · Advanced
Small models next to the data: running inference in every customer's region
Lucas Moreau, Halden
15:20–16:00Studio