Hands-on model selection, tradeoffs and evals w/ Miki Hardisty and Darshaun Nadeau
Frontier models level up constantly — but not every workload needs the newest, biggest, most expensive one. In production, speed, effort, and cost matter as much as raw capability. Prototyping fast? Reach for Opus or Fable. Shipping something that has to be right every time, at volume, on a budget? Reach for GLM, Kimi, Deepseek? That's a different decision, and it should be made with evidence rather than vibes.
This is a technical workshop about replacing model hype with engineering discipline. We'll run real experiments across models live so you can see how they actually compare — and you'll leave with a repeatable method for running the same experiments against your own use cases.
What we'll cover:
Building harnesses so you can swap models without rewriting your stack
Designing scientific experiments: controls, eval sets, and defining what "better" actually means for your workload
Token utilization and the real cost/latency/quality tradeoffs
Recognizing when the smaller, cheaper, faster model is the correct answer
Presented by:
Miki Hardisty
Miki is a local legend and the CEO and co-founder of https://www.olelo-ai.com . She has decades of experience as a technology executive for companies like Jack in the Box.
Darshaun Nadeau
In 1999 he became a market-maker in Japanese domain names, building the largest domain reseller network in the country, then selling his company in 2005 to GMO Internet, today a multibillion-dollar behemoth. He's running the play again from Hawaii as co-founder of FastMetal.ai, an LLM gateway carrying dozens of models. What separates that from a vendor talk: he's also the customer. He runs PSI-Japan (leading corporate ICANN-accredited registrar), Zilarent (a newly founded Hawaii-based AI-first car-rental software with over $2.5M in bookings since 2025), Flowing Blue LLC (Honolulu paid-media agency) and Blue Diamond Vacations (luxury car rental in Waikiki), all running full speed on those same rails. With offices in Honolulu, Tokyo and Hanoi, he's bringing his evals and what it costs him to be wrong.