Building with Jev and System One Models Workshop: 3 Hands On Projects
Hosted by Packt Publishing Limited
View organizer listingThis event has ended. These details are kept for reference.
What to look forward to
Attendees will learn when to use System One models and build three practical projects with Jev: Playwright testing, code triage, and AI observability.
Inside the event
Your AI pipeline is probably sending every decision to an LLM, even the easy ones.
Is this request simple? Is this test result suspicious? Should the agent take the next step?
Each of those calls costs tokens, adds latency and gives an answer your code still has to parse.
Most of these are **System One decisions**: fast, structured and repeatable. They don’t need a model that reasons at length. They need a quick, confident answer your software can act on straight away.
**Jev** is a System One decision model. You give it state and ask a typed question. It returns a structured decision with a confidence score, so your code can branch, rank or route on it directly. When Jev isn’t confident, you hand the case to an LLM. Jev works alongside LLMs rather than replacing them.
In this workshop, you’ll stop guessing and start measuring. You’ll build three working projects that show where System One decisions save money and time, where an LLM is still the right tool, and how to combine them in production.
How the Workshop Runs
This is a hands-on session, not a slide demo. You’ll get the workshop repo, so you can build each project live with Renata or follow along and finish it afterwards with the recording.
Three Projects That Build on Each Other
**Project 1: Cut AI Costs: Measure, Route and Cascade**
You can’t cut what you don’t measure. You’ll start by instrumenting an AI workflow to track token cost, model usage, decision quality and orchestration overhead. Then you’ll put Jev in front of your models so straightforward requests go to a cheaper model and uncertain ones go to a stronger one. To finish, you’ll compare cost and quality before and after, using your own numbers.
**You’ll leave with:** a reusable cost-routing pattern you can add to an existing pipeline.
**Project 2: AI-Powered Exploratory Testing with Jev + Playwright**
Connect Jev to a live application and a Playwright workflow to make fast, structured judgements about application behaviour. You’ll turn observations from an exploratory test into decisions your automation can use immediately, including confidence-aware verdicts and escalation.
**You’ll leave with:** a Playwright workflow that flags suspicious behaviour without a human reviewing every run.
**Project 3: Build a Simple Agent with Jev**
Most agents send every step to an LLM, even when the next move is obvious. You’ll build a simple agent where Jev makes the quick decision at each step and passes only harder or uncertain cases to an LLM. Confidence thresholds decide when the agent acts on its own and when it escalates.
**You’ll leave with:** a working hybrid agent that’s faster, cheaper and more predictable than an LLM-only agent.
What You’ll Be Able to Do
- Tell when a decision needs a System One model and when it needs an LLM
- Measure what your AI workflow actually costs and whether it’s delivering value
- Build a model cascade that sends each request to the cheapest model that can handle it
- Set confidence thresholds that decide whether your software acts or escalates
- Design agents that call an LLM only when they need to
Meet Your Instructor
- **Renata Andrade**
- Quality Engineer Lead | Test Automation Engineer & Architect
Renata helps engineering teams build quality into their software in the fastest and most reliable way possible. She loves sharing practical techniques that teams can use the next day.
Who Should Attend
- **Software developers and AI engineers** building LLM-powered workflows or agents who want lower costs and more predictable behaviour
- **QA and test engineers** exploring AI-assisted testing
- **DevOps, SRE and platform engineers** responsible for AI cost, latency and reliability
- **Engineering leaders** who need measurable results from AI spend
**This workshop may not be for you if** you’re new to programming or haven’t worked with an LLM API before.
What You’ll Receive
- Two-hour live, instructor-led workshop
- Three working projects: AI cost routing, exploratory testing and a hybrid agent
- A reusable repo with all workshop code
- Full workshop recording, so you can rebuild anything you missed
- Workshop slide deck
- Live Q&A with the instructor
- Certificate of completion
###
People on the program
- Renata Andrade
A seat from anywhere.
Register with the organizer for joining instructions and the event link.
Details come from the event listing and the organizer page; registration is managed by the organizer.
Last checked: Nov 16, 2026. Confirm current availability before making plans.
Organizer page checked: Nov 11, 2026. Check registration for the latest details.