Engineering

Cloud & DevOps

Azure and AWS infrastructure for AI and platform workloads: identity, API management, observability, CI/CD and cost control, as code.

At a glance

  • Azure AI Foundry, Azure OpenAI, APIM and Entra set up properly
  • OpenTelemetry and Application Insights from day one
  • CI/CD, containers and infrastructure as code
  • Cost visibility per tenant, per model, per tool call

Momentaj, a Toronto AI engineering firm, builds the Azure and AWS infrastructure that AI and platform workloads run on. AI workloads change what good infrastructure looks like: model endpoints with quotas, token costs that need accounting, MCP servers that need OAuth and gateways, and traces that have to follow a request through several model calls. We set that up, and the ordinary cloud work around it.

What we build

  • Azure for AI. Azure OpenAI and Azure AI Foundry deployments, API Management in front of MCP servers and APIs, Entra and Entra External ID for identity, private networking and key management.
  • AWS and hybrid. Where your estate already lives on AWS or on-premises, including private clouds for data that cannot leave.
  • Observability. OpenTelemetry tracing across services and model calls, Application Insights or Grafana dashboards, alerts that mean something.
  • Delivery. CI/CD pipelines, Docker and Kubernetes where warranted, infrastructure as code, environments that match production.
  • Cost control. Token and compute accounting per tenant, model and tool call, with budgets and alerts.

How it's done

Start from the workload and its data rules, write the infrastructure as code, and make the first deployment observable before the second feature ships. Azure DevOps or GitHub Actions for delivery.

Proof

Momentrack: a seven-service private cloud on our own hardware, with every link measured. uBeac: a multi-tenant platform ingesting 100 million documents a week on MongoDB and RabbitMQ.

Frequently asked questions

Can you work inside our Azure tenant?

Yes, and we prefer it. The infrastructure, keys and data stay yours; we work through your subscription with the access you grant.

Do we need Kubernetes to run AI workloads?

Usually not at the start. Container apps, app services or plain VMs cover most workloads with far less to operate. We move to Kubernetes when scale or isolation needs justify it.

Which clouds do you work with?

Azure most often, and AWS. Many of our clients already run on Microsoft, so Azure OpenAI and AI Foundry, API Management, Entra and Application Insights are the usual pieces.

How do you keep AI costs under control?

By measuring them where they happen: per tenant, per model and per tool call, with budgets per request and per month and an alert before the bill arrives.

Do you keep cloud infrastructure as code?

Yes. Infrastructure, CI/CD and configuration are defined as code and reviewed like code, so an environment can be rebuilt and every change has a history.

Do you work with companies outside Toronto?

Yes. We work remotely with clients across Canada, the United States, the Middle East and beyond, and in person across the Greater Toronto Area.

Related services

IoT & Connected Systems

Devices, radios and the cloud that receives their data: BLE, LoRa and LTE-M hardware, MQTT ingestion, real-time dashboards and the analytics on top.

Web & Mobile Development

Fast, accessible web apps and mobile-first PWAs in Blazor, SvelteKit and React, designed with the people who will use them.

Custom Software Development

Platforms, back offices and integrations built to run for years: .NET and Python services, queues, payments, real-time delivery, and the APIs an AI system will need.

Talk to the people who would build it

Tell us what you are trying to build. We will come back with an honest take and a plan.

Book an AI consultation