Presented by
WideOps Google Cloud

AI Agents in Production on Google Cloud

Two talks on running AI agents at scale

A free, in-person morning at Google’s Cambridge office. WideOps’ CTO and Principal Architect share what really happens when agent fleets hit rate limits, and how to make agents prove their answers.

Register now

Free to attend. Seating is limited.

Yuval Mizrahi presenting a Google Cloud session to a room of attendees
Yuval Mizrahi at a recent WideOps session hosted at Google.
Date
Thursday, November 12, 2026
Time
9:00 AM – 1:30 PM ET
Hosted by
WideOps & Google Cloud

Agents shine in demos. Production is different.

Agents don’t make one call per request. They loop, retry, and share the same limits, and that changes both your cloud bill and how far you can trust their answers.

The first session shows how a rate limit that costs nothing can turn into a 50× jump in model spend, and the guardrails that keep agent fleets predictable.

The second shows why agents are often wrong in a convincing way, and how to make them back every conclusion with a real test.

What you’ll learn

  1. 01
    Keep agent costs predictable

    Stop retry loops and runaway token use early.

  2. 02
    Put guardrails in the right place

    Limits in the agent, gateways in front of tools and models.

  3. 03
    Make agents prove their answers

    Trust a conclusion only when it passes a real test.

  4. 04
    Know when something goes wrong

    Spot loops within minutes, and catch features that fail silently.

Format
2 technical sessions
Duration
Half day, in person
Level
Intermediate
Admission
Free

The sessions

Two talks based on real production systems, outages, and cloud bills, with time for questions.

  1. 01

    10,000 Agents Walk Into a Rate Limit

    • Why a free error can cost $50,000
    • Three war stories from production
    • Three layers of guardrails
    • Catching it before the invoice

    How agent fleets turn rate limits into runaway cloud bills, and a checklist to prevent it.Yuval Mizrahi, Principal Architect

  2. 02

    Your Agent Is Lying to You, and It Sounds Great

    • Conclusions that must pass a test
    • Showing what was never verified
    • Memory that learns what helped
    • When safety nets fail silently

    Why a fluent answer is not proof, and how to make agents back every conclusion with evidence.Guy Shafir, CTO

Google Cloud products covered

Vertex AI and Gemini, the Agent Development Kit (ADK), Spanner Graph, Cloud Run, Pub/Sub, and Cloud Monitoring.

What to expect

Real production stories and architecture, not product pitches. There is no hands-on coding.

Agenda

Thursday, November 12, 2026. All times are Eastern Time.

  1. 9:00 – 9:30 AM

    Breakfast, coffee & networking

  2. 9:30 – 10:00 AM

    Welcome & introduction to WideOps US

    Elad KatavGeneral Manager, WideOps USA

    Guy ShafirCTO, WideOps

  3. 10:00 – 10:45 AM

    10,000 Agents Walk Into a Rate Limit

    • Why agent costs spike
    • Guardrails for tools and models
    • Alerts before the invoice

    Yuval MizrahiPrincipal Architect, WideOps

  4. 10:45 – 11:05 AM

    Break

  5. 11:05 – 11:50 AM

    Your Agent Is Lying to You, and It Sounds Great

    • Testing agent conclusions
    • Honest investigation reports
    • Memory that learns

    Guy ShafirCTO, WideOps

  6. 11:50 AM – 1:30 PM

    Q&A, next steps & networking

Speakers

The WideOps team presenting at Google Cambridge.

1:1 advisory sessions

Want to go deeper on your own setup? Request a private session with one of our Google Cloud experts on the day of the event. Availability is limited, and we confirm each session by email.

Request a session

Bring a real question, for example

  • Reviewing your AI agent architecture
  • Controlling Vertex AI and Gemini costs
  • Making agent answers verifiable
  • Google Cloud architecture for AI workloads

Who should attend

Teams that run AI agents in production, or plan to soon.

Engineering & technology leaders

Scale AI agents without surprise bills or bad answers.

Cloud & solution architects

Design the limits and checks that keep agents reliable.

Platform, SRE & DevOps teams

Protect shared systems and catch failures before users do.

AI & ML engineers

Apply proven patterns for budgets, retries, and testing.

Recommended background: familiarity with LLMs, AI agents, or cloud infrastructure. The sessions are technical, but no coding is involved.

Register for the event

Attendance is free. Seating is limited. Submit your details and we’ll confirm your registration by email.

Admission
FreeBreakfast is included. There is no registration fee.
Date
Thursday, November 12, 2026
Time
9:00 AM – 1:30 PM ETBreakfast and networking from 9:00 AM
Location
Google Cambridge, 4CC, 90 Broadway St, Cambridge, MA 02142Open the office in Google Maps.
What to bring
Government-issued photo ID for building access.
1:1 advisory
Optional. Request a session in the form.

Your details

All fields are required unless marked optional.

Please enter your first name.
Please enter your last name.
Please enter a valid work email.
Please enter your company.
Please enter your job title.
Please select your area of expertise.
Please enter a valid US phone number.
Please answer “What’s your biggest challenge with AI agents today?”.
Please accept to continue.

Your information is shared only with WideOps and Google Cloud for this event.

Frequently asked questions

Have another question? Email us at info@wideops.com.

Yes. Attendance is free, including breakfast. Seating is limited, and each registration is confirmed by email.

They are technical and based on real production systems, but accessible to technology leaders. There is no hands-on coding.

Yes. When you register, check the 1:1 advisory option and add a topic if you have one, so we can match you with the right WideOps expert. Sessions take place on the day of the event, and we confirm your time by email.

A government-issued photo ID for building check-in. A laptop is optional.

The event is at Google Cambridge, 4CC, 90 Broadway St, Cambridge, MA 02142, near the Kendall/MIT MBTA station. Open Google Cambridge in Google Maps. Check-in instructions are sent with your confirmation email.

No. The sessions include candid production stories, so they are in person only and will not be recorded or streamed.

Yes. Each attendee needs to register individually so we can arrange building access.

November 12 · 90 Broadway9:00 AM – 1:30 PM ET Register