At AWS User Group Toronto, community feedback is at the heart of what we do.
After every event, we listen carefully to what our attendees want to learn next. Our mission is to help cloud enthusiasts, builders, and tech aspirants grow their skills by bringing the most relevant and in-demand topics to our meetups.

One common theme we kept hearing was the strong interest in Kubernetes optimization—along with a desire to explore new AWS services and AI-driven customer experience solutions.

We heard you.

Thus, on April 24, 2025, we hosted a packed evening focused on two highly requested topics:

  1. Automated Kubernetes Resource Optimization

  2. Future of Customer Experience with Amazon Connect and Amazon Q

It was a night full of technical insights, real-world demos, and valuable community connections.
Here’s a full recap of everything we covered.

🔹 Session 1: Automated Kubernetes Resource Optimization

Speaker: Andrew Hillier, CTO & Co-founder @ Densify

Overview:
Andrew Hillier opened the evening by addressing a foundational challenge for Kubernetes operations: how to manage CPU and memory resources effectively.
In Kubernetes, resources have a direct impact on application performance, environment stability, and operational costs.

While Kubernetes is flexible, Andrew highlighted that most organizations either over-provision (leading to unnecessary cost) or under-provision (causing instability and downtime)—simply because managing resources manually is extremely difficult at scale.

Key Topics Covered (Explained):

  1. Why Resource Optimization Matters
    Resources like CPU and memory are fundamental to container performance:

    • Under-provisioned workloads cause throttling, latency, and potential crashes.
    • Over-provisioned workloads waste infrastructure, inflating cloud bills without any benefit. Optimizing resources correctly leads to better system reliability, customer experience, and cost control.

  2. Why Manual Optimization Fails
    Andrew explained why relying on human judgment for setting resource limits and requests isn't sustainable:

    • Workloads constantly fluctuate based on time, season, and user behavior.
    • Microservices have complex interdependencies; a performance issue in one service can cascade.
    • Developers often overestimate needs to “be safe,” resulting in systematically inflated infrastructure costs.
    • Without continuous analysis and adjustment, even well-designed systems become inefficient over time.

  3. Intelligent Automation: The Next Step
    True optimization automation needs deep analytics, not just reactive scaling:

    • Historical workload analysis is required to predict safe, efficient resource levels.
    • Correlation across services ensures dependencies are respected.
    • Recommendations must be actionable at scale, often via GitOps pipelines or API-driven updates to Kubernetes manifests and Helm charts.

The goal is continuous optimization, where cluster efficiency improves without sacrificing performance or stability.

Live Demo Summary:
Andrew demonstrated:

  • Real workload telemetry capture.
  • Analysis using machine learning models to propose new CPU and memory settings.
  • GitOps-based application of these recommendations in a safe, repeatable manner.

The result: reduced cloud costs, higher reliability, and minimal manual intervention.

Audience Questions:
Key questions during the session included:

  • How to handle stateful services during optimization?
  • What strategies avoid restart storms when changing pod resources?
  • How to integrate optimization with DevOps and CI/CD workflows?

🔹 Session 2: Future of Customer Experience with Amazon Connect and Amazon Q

Speaker: Manush Parikh, Manager, Solutions Architect @ ScaleCapacity

Overview:
Shifting focus to customer engagement, Manush Parikh introduced how businesses can deliver faster, smarter, and more personalized customer experiences using Amazon Connect and Amazon Q.

Modern customers demand seamless support, and AWS has redefined what’s possible for cloud contact centers by integrating scalable infrastructure with AI-driven intelligence.

Key Topics Covered (Explained):

  1. Amazon Connect: Redefining the Contact Center
    Manush explained how Amazon Connect provides:

    • Rapid deployment: Set up a contact center within hours, without heavy engineering.
    • Omnichannel support: Manage voice, chat, SMS, video—on a unified platform.
    • Real-time analytics: Monitor performance, wait times, agent productivity in real-time.
    • Unified agent workspaces: Enable agents to have complete customer context during interactions.
    • Workforce management tools: Forecast demand, schedule agents, optimize staffing.

  2. Amazon Q in Connect: Generative AI for Smarter Service
    Manush introduced Amazon Q as a revolutionary enhancement to Connect:

  • Real-Time Assistance:

    • Q analyzes conversations live and recommends responses, articles, or actions.

  • Knowledge Base Integration:

    • Pulls answers from organizational FAQs, policy docs, and support articles.

  • Personalized Interactions:

    • Tailors answers based on customer history, membership status, and active case data.

  • Guardrails and Governance:

    • Admins can control Q's tone, information access, and restrict hallucinations.

By embedding Amazon Q, businesses reduce agent load, boost response quality, and achieve higher first-contact resolution rates.

Live Demo Summary:
Manush presented a real-world use case:

  • A customer called about canceling a rental booking.
  • Amazon Q instantly pulled cancellation policies.
  • It proposed a detailed, personalized response for the agent.
  • It suggested potential upsell options (e.g., rescheduling instead of canceling).

The flow highlighted AI-assisted resolution paths, personalized service, and reduced average handling time.

Audience Questions:
Manush fielded questions like:

  • How is data privacy enforced when Amazon Q accesses internal knowledge bases?
  • How do you audit and supervise AI-generated responses?
  • What strategies allow scaling Amazon Connect for multi-brand, multi-region operations?

He emphasized strict access controls, compliance-friendly architecture, and tiered knowledge bases for complex organizations.

👥 Community Networking
The evening concluded with lively discussions among developers, architects, DevOps engineers, and CX professionals.

Themes like optimizing Kubernetes clusters, embedding AI in customer service, and automating operations safely dominated conversations.

🔗 Stay Connected
We are building an inclusive, technically focused AWS community here in Toronto. If you're passionate about cloud-native technologies, AI, and innovation, join us:

🔗 AWS User Group Toronto on LinkedIn
📅 AWS UG Toronto on Meetup

💬 Your Turn
Have you implemented Kubernetes optimization strategies?
Are you starting to use AI assistants like Amazon Q ?
We would love to hear about your experiences—drop a comment below!

—
Bansi Delwadia
AWS User Group Toronto Leader | AWS Community Builder