Cloud platform and DevOps engineering: one platform for apps and AI workloads
The service builds a repeatable cloud platform that carries both ordinary applications and AI workloads: infrastructure as code, CI/CD and GitOps, inference hosting with GPU or queue capacity, observability, backup and disaster recovery, secrets and identity management, and the cost controls that keep it all accountable. The platform is handed over to your team, or moved into Managed Operations, once it is running.
Submit a project Assess an inference workload firstReviewed by David Nguyen (CEO) · Updated 29 Sep 2026
OutsourcingVN is operated by Netbase JSC. Inference hosting and GPU capacity are new operating questions for every supplier in the current AI era; Netbase takes them on and combines them into the same platform discipline that already runs ordinary application infrastructure, rather than bolting them on as a separate project.
What does the platform build?
Environments defined in version-controlled code, reproducible from a clean account rather than assembled by hand.
Build, test and deployment pipelines with promotion between environments and a rollback path for every release.
Serving infrastructure for AI workloads sized on real load, alongside the application infrastructure it supports.
Metrics, logs and traces wired to alerts that reach a named owner, covering both application journeys and inference endpoints.
Backups that are tested by restoring them, not only scheduled, with a recovery time a named owner has agreed to.
Credentials held in a secret store, role-based access and rotation on a schedule rather than on discovery of a leak.
Budgets, alerts and resource tagging so spend on compute and inference capacity is visible before the invoice arrives.
Good fit and poor fit
Good fit
- Applications and AI workloads both need a repeatable platform, not a one-off environment
- Infrastructure as code, CI/CD, observability or cost controls are missing or inconsistent across environments
- GPU or queue capacity for AI inference needs sizing, hosting and monitoring alongside ordinary application infrastructure
- A named platform owner can accept infrastructure as code, secrets and IAM, and backup and DR as delivered, then hand over or move to managed operations
Another route fits better
- The system is already built and just needs day-to-day operation; Managed Operations is the ongoing route
- Whether an inference workload is even ready for production is still open; the production AI inference assessment settles that first
- The question is reliability and incident readiness on an existing system rather than building the platform; the reliability and incident readiness guide covers that assessment
- Access, data location and vendor questions are still open; the data security and compliance guide works through those first
How the platform is built
-
Map the current infrastructure
Inventory environments, deployment paths, existing pipelines, dependencies and the workloads, application and AI, the platform must carry.
-
Build infrastructure as code and CI/CD
Stand up version-controlled environments and GitOps pipelines with promotion, testing gates and a rollback path.
-
Wire inference hosting and observability
Size and deploy GPU or queue capacity for AI workloads, then connect metrics, logs, traces and alerts to a named owner.
-
Harden and hand over
Complete backup and DR testing, secrets and IAM, and cost controls, then hand the platform to your team or move it into Managed Operations.
Netbase's delivery lifecycle runs discovery and strategic alignment, team assembly and architecture planning, agile execution with outcome-based milestones, modular components, training and rollout, and ongoing support; platform work follows the same shape, with infrastructure treated as a milestone deliverable rather than an afterthought.
Where AI changes the platform
An inference endpoint behaves differently from an ordinary web request: it needs GPU or accelerator capacity, longer and more variable response times, and cost that scales with tokens rather than requests. This service treats that as one more workload the platform must host, monitored and capped like any other, not a separate infrastructure track. Where inference has to run on a device rather than in the cloud, the operating case is different again; the edge AI deployment assessment guide covers offline behaviour, device lifecycle and field rollback for that scenario.
A worked scenario: an inference workload added to an existing platform
A logistics company runs its core application on infrastructure that was never fully defined as code, and now wants a document-extraction model to process delivery paperwork overnight.
The platform engagement starts by converting the existing environments to infrastructure as code so both the current application and the new workload can be reproduced and reviewed the same way. A GitOps pipeline is built for both, with the extraction workload's batch nature (a nightly job with a morning deadline) captured as a queue with retry and idempotency rather than treated as a live endpoint. Observability is extended to cover job success rate, queue depth and processing time, alerting a named operations owner when the overnight run is late. Cost controls cap GPU spend per run and alert if a batch runs long. The platform is accepted against those observability and cost targets, then handed to Managed Operations for ongoing running.
What technical stack does Netbase draw on?
Since 2020 Netbase has worked as offshore development and managing partner on a multi-tenant cloud ERP SaaS platform for a US client that is not named. Its first phase, from 2020 to 2023, covered CRM, real-time messaging, HR, a knowledge base, custom fields and workflows, work and project management and API integrations, on a stack that includes React and Next.js, Laravel and Strapi, PostgreSQL, MySQL, MariaDB and MongoDB on AWS; a later retail phase was planned work when that profile was written and is not presented as delivered. AWS is the cloud Netbase already runs a production, multi-year platform on, which is the practical base this service draws its infrastructure patterns from.
See the multi-tenant cloud ERP SaaS platform record
What company-level assurance backs this service?
Netbase JSC holds ISO 27001 certification for information security management, and Netbase JSC holds a SOC 2 Type II attestation; each describes Netbase's own organisation and is never extended to a client's infrastructure or hosting. Security practices include secure code review and version control, role-based access control, multi-factor authentication for admin dashboards, contributors under NDA, and NDAs and data processing agreements available on request, applied to the platform's secrets and IAM design as much as to application code.
Evidence and related work
For an existing client store, Netbase scoped and delivered a three-phase recovery of a compromised Magento site: investigation with an infected-file report and recovery plan, cleaning and system recovery, then file restore, recoding and server hardening, typically 6 to 13 days depending on the damage. The Magento recovery and hardening record is the closest published example of the server-hardening discipline this service builds proactively, before an incident forces it. The multi-tenant cloud ERP SaaS platform record shows the AWS-hosted, multi-year infrastructure base above.
Since 2020, Netbase has worked as offshore development and managing partner on a multi-tenant cloud ERP that a US software company offers as SaaS to small and mid-sized businesses.
Keep Reading
For an existing client store running Magento that had been compromised, Netbase scoped a three-phase recovery: investigate and assess, clean and restore, then repair and harden.
Keep ReadingQuestions to ask a cloud platform supplier
-
Is our infrastructure defined as code today, or only partly?
An honest gap list, not a claim everything already is
-
How is GPU or queue capacity for inference sized?
A load test on real request or job shapes, not a vendor benchmark
-
When was a backup last restored, and how long did it take?
A dated test and a recorded duration
-
Who is alerted when a deployment, job or inference endpoint fails?
A named owner and a tested alert route
-
What happens to unused GPU capacity?
A cost control that caps or scales it down, not an unmonitored bill
-
Does this hand over to our team or to a managed service?
A stated choice, agreed before the platform is built
What failure modes should you watch for?
-
Infrastructure drifts from its code
Signal: a change made by hand in the console does not match the repository. Owner: the platform owner, who enforces changes through the pipeline only.
-
Inference capacity is sized on a demo, not real load
Signal: the endpoint slows or errors once real traffic arrives. Owner: the platform owner, who load-tests before rollout.
-
Backups exist but are never restored
Signal: a restore is attempted for the first time during an outage. Owner: the platform owner, who rehearses restores on a schedule.
-
Secrets sit outside the secret store
Signal: a credential is found in a repository or a chat message. Owner: the security owner, who rotates it and closes the gap.
-
Cost is discovered on the invoice
Signal: a GPU bill surprises the finance owner. Owner: the platform owner, who wires budget alerts before launch.
Frequently asked questions
This service builds the platform: infrastructure as code, CI/CD, inference hosting, observability and cost controls. Managed Operations runs a system that is already accepted, on terms agreed in its own assessment. Many engagements build here and move to managed operations once the platform is live.
Assess the workload first. The production AI inference assessment settles the workload boundaries, serving route and acceptance cases; this service then builds the platform that hosts the accepted design.
Yes. Infrastructure as code, secrets and IAM are scoped to run inside your account, with access granted per role rather than through a shared credential.
Not as a published default. The Hanoi office works Monday to Saturday, 9:00-18:15 Vietnam time (UTC+7), and support coverage follows those days; wider coverage is agreed per engagement, most often once the platform moves into Managed Operations.
Cost controls cap and can scale down idle capacity, with usage reported to a named owner rather than discovered on the bill.
Not directly. Cloud platform engineering hosts inference in your cloud or account; where a model must run on a device in the field, the edge AI deployment assessment covers that separate operating case.
Ready to line up the platform?
Bring the workloads the platform must carry, what is already defined as code versus assembled by hand, and whether the goal is handover to your team or a move into managed operations. Submit a project with that scope, and a person will reply with which part to build first. OutsourcingVN is Netbase's own outsourcing-services platform.