What DevOps consulting from India looks like for a Singapore company
A remote DevOps consultancy for a Singapore company should work inside your own AWS, Azure or Google Cloud accounts in the Singapore region, overlap at least six hours with Singapore Time (2.5 hours ahead of India), and deliver pipelines and Terraform that already produce the evidence PDPA and, for financial firms, MAS’s Technology Risk Management Guidelines expect: reviewed changes, separated release duties, MFA on privileged access and tested recovery.
Most searches for DevOps in Singapore end on a job board
Search for DevOps consulting in Singapore and most of page one is job listings, with a couple of agency pages that could be about any city. That tells you something about the market: Singapore companies mostly try to hire DevOps engineers first, find the role slow and expensive to fill, and only then look for a team that can do the work while the hire happens, or instead of it.
We are that kind of team, with clear limits. Radiatus has no Singapore entity and no staff in Singapore. Our engineers work from India, 2.5 hours behind Singapore Time, inside your cloud accounts and your Git organisation, with named access that you grant and revoke. We do not resell cloud, hold no partner status with AWS, Microsoft or Google, and do not certify anything. What we bring is the engineering: CI/CD that your auditors can follow, Terraform or OpenTofu that rebuilds an environment from scratch, and Kubernetes clusters that are upgraded on schedule rather than when the provider forces it.
Our general approach is on the DevOps automation page, and cluster design lives on Kubernetes consulting. This page covers what changes when the client is in Singapore: the time-zone split, the regions, PDPA, and the MAS rules that shape how a regulated firm may let an outside team touch production.
What is different about DevOps for a Singapore company
The time-zone gap is small and useful. Singapore runs on UTC+8 and India on UTC+5:30, with no daylight saving on either side, so Singapore is always 2.5 hours ahead. A conventional Indian working day of 09:30 to 18:30 IST is 12:00 to 21:00 SGT. Most Singapore clients prefer the Indian day shifted earlier, which gives six to eight shared hours. The quiet hours that suit production changes in Singapore, around midnight to 03:00 SGT, fall at 21:30 to 00:30 IST: late evening for us, not the middle of the night.
All three clouds run a Singapore region with three zones. AWS Asia Pacific (Singapore) is ap-southeast-1 with three Availability Zones, Microsoft Azure’s Singapore region is called Southeast Asia (southeastasia) with three availability zones, and Google Cloud’s is asia-southeast1. One detail catches teams out: Azure pairs Southeast Asia with East Asia in Hong Kong, so geo-redundant storage replicates there by default. If your contracts or your regulator expect data to stay in Singapore, use zone-redundant options instead and record the decision.
Nearby regions widen disaster-recovery choices. AWS now runs regions in Malaysia, Jakarta and Thailand as well as Mumbai and Hyderabad, and Azure has Malaysia West in Kuala Lumpur. That makes a second-region recovery plan practical, but every copy outside Singapore is a transfer of personal data under the PDPA and possibly an outsourcing question for an MAS-regulated firm. We document where each backup and log stream lands so your data protection officer can sign it off.
Regulation is about evidence, not tools. The PDPA applies to every organisation handling personal data in Singapore. MAS’s Technology Risk Management Guidelines, issued on 18 January 2021 and still the current version, apply to banks, insurers, capital-markets and payment firms, with binding notices on technology risk and cyber hygiene reissued under the Financial Services and Markets Act on 9 May 2024. None of them name a CI server or a cloud. They ask you to show that changes are reviewed and reversible, that no one person can push code to production alone, and that privileged and remote access is controlled. That is pipeline design work.
Hiring is slow, which is why people look offshore. A foreign DevOps engineer needs an Employment Pass, and the Ministry of Manpower’s qualifying salary is S$5,600 a month for new applicants outside financial services and S$6,200 in financial services, rising with age. From 1 January 2027 those floors rise to S$6,000 and S$6,600 for new applications, according to MOM’s eligibility page. Those are minimums, not market salaries, and the COMPASS points test applies on top.
Working hours: what 2.5 hours actually means
The gap is short enough for one shared working day and long enough to give Singapore an early-evening hand-off. This is the pattern we agree most often; the exact hours are written into each engagement.
| Singapore (SGT, UTC+8) | India (IST, UTC+5:30) | What usually happens |
|---|---|---|
| 10:00 | 07:30 | Shifted Indian day starts; daily stand-up with your team |
| 10:00 – 18:00 | 07:30 – 15:30 | Shared hours: pairing, reviews, planned changes in non-production |
| 18:00 – 19:00 | 15:30 – 16:30 | Written hand-off: what changed, what is waiting for your approval |
| 00:00 – 03:00 | 21:30 – 00:30 | Production change windows, scheduled in advance with your approver |
| Overnight | Evening and night | Out-of-hours on-call only if you buy it as a separate, written arrangement |
The part that needs agreement is approval. If MAS rules or your own policy require a Singapore staff member to approve production changes, the change window has to fit their hours too, so we plan windows a week ahead rather than asking for approvals at midnight.
Choosing the region and where copies go
| Cloud | Singapore region | Zones | Watch out for |
|---|---|---|---|
| AWS | Asia Pacific (Singapore), ap-southeast-1 | 3 | Cross-region backup copies and replication are opt-in; check every AWS Backup plan and S3 replication rule for a second region |
| Microsoft Azure | Southeast Asia, southeastasia | 3 | The paired region is East Asia (Hong Kong), so geo-redundant storage and some recovery features copy data there unless you choose zone-redundant options |
| Google Cloud | Singapore, asia-southeast1 | 3 | Multi-region bucket locations such as ASIA spread data across countries; use the regional location if data must stay put |
Sources: AWS Regions list, Azure regions list and Google Cloud locations, checked 13 November 2026. For most Singapore workloads, three zones in one region already meet the availability you need; a second region is for disasters, and moves the conversation into data transfer.
PDPA, translated into platform controls
This is engineering guidance, not legal advice. Your data protection officer or counsel decides what applies; we build the controls and the records that show they work.
| PDPA obligation | What it means for the platform | What we typically set up |
|---|---|---|
| Protection | Reasonable security arrangements for personal data you hold or control | Encryption at rest with keys you own, private networking for databases, least-privilege roles, MFA for every human, central audit logs |
| Retention limitation | Stop keeping personal data once the purpose is served | Lifecycle rules on buckets and log groups, backup retention set per data class, a register of what each store holds |
| Transfer limitation (section 26) | Personal data sent outside Singapore needs a comparable standard of protection, and your organisation stays responsible when a provider does the sending | Production data kept in the Singapore region, masked data in non-production, remote sessions that view rather than copy data, and contract terms you can show |
| Data breach notification (sections 26C and 26D) | A provider processing data for you must tell you without undue delay; you must notify the PDPC within three calendar days of assessing that a breach is notifiable | Alerting on unusual access, a runbook with named contacts on both sides, and logs that let you assess scope quickly |
Since 1 October 2022 the maximum financial penalty under the PDPA has been 10% of annual turnover in Singapore for organisations with more than S$10 million in local turnover, or S$1 million otherwise (section 48J; see this summary by Allen & Gledhill). Remote access by an engineer in India to systems holding personal data can itself raise transfer questions; we prefer designs where our engineers manage infrastructure and pipelines without needing to read production records at all.
MAS technology-risk rules, translated into pipeline controls
The Technology Risk Management Guidelines are principles-based. The notices are binding on the institutions they name. Here is how the clauses we meet most often become configuration.
| Source | What it asks | How it shows up in the pipeline |
|---|---|---|
| TRM Guidelines 6.3 (DevSecOps) | DevSecOps aligned with change, configuration and release management, with segregation of duties across development, testing and release | Protected branches, required reviewers who are not the author, deploy jobs that only run from the main branch |
| TRM 7.5 (change management) | Changes assessed, tested and approved; a backup and rollback plan; logging enabled during the change; defined emergency-change procedure | Pull request as the change record, Terraform plan attached, automatic rollback on failed health checks, a documented break-glass path |
| TRM 7.6 (software release) | No single individual able to develop, compile and move code between environments; traceability and integrity of code moved | Builds by the CI system only, images signed and promoted by digest, approval by a second person before production |
| TRM 9.2 and 9.3 (privileged and remote access) | Privileged access on a need-to-use basis, logged and reviewed; MFA and secured devices for remote access | Just-in-time elevation through AWS IAM Identity Center or Entra PIM, session logging through Session Manager or Bastion, no shared admin users |
| TRM 3.5 (competency and background review) | Background checks for personnel, including contractors, who access the institution’s data and systems | Not a pipeline control: ask any provider, including us, what checks they can evidence for named engineers |
| MAS notices on technology risk (for example FSM-N13) | Critical systems: unscheduled downtime no more than 4 hours in 12 months, recovery time objective of 4 hours or less, MAS notified within 1 hour of discovering a relevant incident, root-cause report within 14 days | Multi-zone designs, restore tests timed against the 4-hour objective, alert routes that reach your incident manager within minutes |
| MAS notices on cyber hygiene (for example FSM-N06 for banks) | Admin accounts secured; MFA for admin accounts on critical systems; patches applied in a risk-based timeframe; written security standards for every system | MFA enforced in the identity provider, patch baselines for nodes and images, CIS-style benchmarks written down per system type |
Sources: MAS Technology Risk Management Guidelines (January 2021), and MAS notices FSM-N13 and FSM-N06, both issued on 9 May 2024. Banks should also read MAS notice 658 and the revised outsourcing guidelines, in force since 11 December 2024; whether our work counts as an outsourced relevant service for you is your compliance team’s decision, and we will supply the due-diligence answers they need.
What segregation of duties looks like in a real workflow
A trimmed example of a production deploy job in GitHub Actions. The production environment is configured with required reviewers and “prevent self-review”, so the person who merged cannot also approve the release; the job gets short-lived AWS credentials through OIDC instead of stored keys, and it refuses to deploy an image whose signature does not verify.
name: deploy-production
on:
workflow_dispatch:
inputs:
digest: { description: "sha256 digest from the build job", required: true }
permissions:
id-token: write # OIDC token for AWS; no long-lived keys in the repo
contents: read
env:
IMAGE: 111122223333.dkr.ecr.ap-southeast-1.amazonaws.com/payments-api
jobs:
deploy:
runs-on: ubuntu-latest
environment: production # required reviewers, self-review blocked
steps:
- uses: actions/checkout@<pinned-commit-sha>
- uses: aws-actions/configure-aws-credentials@<pinned-commit-sha>
with:
role-to-assume: arn:aws:iam::111122223333:role/gha-deploy-prod
aws-region: ap-southeast-1
- run: cosign verify --key cosign.pub "$IMAGE@${{ inputs.digest }}"
- run: |
aws eks update-kubeconfig --name prod-sg
helm upgrade payments-api ./chart --atomic --timeout 10m \
--set image.digest=${{ inputs.digest }}
The run log, the approver’s name, the pull request and the signed digest together form the change record that TRM 7.5 and 7.6 describe. Nobody types it up afterwards.
Kubernetes consulting for Singapore: the upgrade calendar
Most of the Kubernetes work Singapore teams bring us is not design but catching up. On Amazon EKS, a minor version gets 14 months of standard support and then 12 months of extended support, which costs six times the standard control-plane rate per cluster-hour, according to the EKS pricing page. Dates from the EKS version calendar as of 13 November 2026:
| Version | End of standard support | End of extended support | What to do now |
|---|---|---|---|
| 1.31 | 26 Nov 2025 | 26 Nov 2026 | The control plane is upgraded automatically after this date; node groups and add-ons are not, so upgrade them yourself first |
| 1.32 and 1.33 | Passed | 23 Mar 2027 and 29 Jul 2027 | Already paying the extended rate; plan the climb to a standard-support version |
| 1.34 | 2 Dec 2026 | 2 Dec 2027 | Upgrade before 2 December to avoid the extended rate |
| 1.35 to 1.37 | 27 Mar 2027 to 1 Dec 2027 | 2028 | In standard support; keep a rhythm of one minor every four months or so |
AKS and GKE have their own long-support options and price them separately. Whatever the cloud, the work is the same: scan manifests and Helm charts for removed APIs, upgrade add-ons, rehearse on a staging cluster, then move control plane and nodes one minor version at a time. Our Kubernetes consulting page covers the method and when not to use Kubernetes at all.
Hire in Singapore, use a local consultancy, or retain an offshore DevOps team?
| Option | Public cost reference | Strengths | Trade-offs |
|---|---|---|---|
| Hire a DevOps engineer in Singapore | For an Employment Pass holder, MOM’s qualifying salary floor is S$5,600 a month (S$6,200 in financial services) for younger candidates, up to S$10,700 (S$11,800) at 45 and above; S$6,000 to S$11,500 (S$6,600 to S$12,700) for new applications from 1 Jan 2027 | Full-time ownership, on site, same time zone | Floors are not market rates; hiring takes months; one person is a single point of failure |
| Singapore consultancy | Day rates are not published consistently; ask for them in writing | On-site workshops, local entity, familiar with local procurement | Usually the most expensive per day; check who actually does the work |
| Offshore DevOps team in India (what we offer) | Quoted in engineer-days after a review; see the ranges below | Several skills on one retainer, overlap of six to eight hours, continuity when one person is away | No on-site presence; data-transfer and outsourcing questions must be settled first |
How much effort the common pieces take
These are our planning ranges in engineer-days for a typical Singapore SaaS or fintech estate. They are not a quote: the review sets the scope, and the price follows from it.
| Work | Typical effort | What moves the number |
|---|---|---|
| DevOps and platform review, one product on one cloud | 5 – 8 engineer-days | Number of repositories, environments and accounts |
| CI/CD with approvals, OIDC and signed images for one service family | 10 – 20 | How many services, and whether tests already exist |
| Terraform or OpenTofu for an account built by hand | 15 – 35 | Resource count, drift, how many environments to import |
| One Kubernetes minor-version upgrade, per cluster | 2 – 5 | Add-ons, removed APIs, stateful workloads |
| Catching up a cluster three or more versions behind | 8 – 20 | Whether a blue-green rebuild beats in-place steps |
| MAS-style change and access evidence across existing pipelines | 8 – 15 | Number of pipelines and identity providers |
| Retained offshore DevOps team | Agreed monthly block of engineer-days | On-call needs, release frequency, number of clusters |
Questions to ask any DevOps consultancy serving Singapore
- Where are the engineers who will hold production access, and which entity employs them?
- Will you work only inside our accounts and Git organisation, with named users we can revoke?
- How will you avoid copying personal data out of the Singapore region?
- Which of your deploy steps enforce segregation of duties, and can we see a sample run log?
- What background checks can you evidence for the named engineers?
- How quickly will you tell us about a suspected breach, and is that in the contract?
- Which Kubernetes versions are our clusters on, and what does each month of extended support cost us?
- What do we get at handover, and could our own team rebuild the platform from it?
DevOps and Kubernetes consulting services for Singapore teams
Four pieces of work. Most Singapore clients start with the review and one build item, then decide whether to keep a retained team.
DevOps and platform review
The problem
Releases depend on one person, nobody is sure what Terraform would do if run today, and a customer or regulator has started asking how changes reach production.
How we do it
Read-only access to your cloud accounts, Git and CI for a week. We map the release path, identity and secrets handling, backups and cluster versions, then rank what to fix by risk, including the MAS or PDPA items your team has flagged.
What you get
- Release-path map from commit to production
- Ranked findings with effort in engineer-days
- Region and data-location record
- Kubernetes version and extended-support exposure
CI/CD and infrastructure as code
The problem
Deploys are scripts on a laptop, environments differ in ways nobody can explain, and long-lived cloud keys sit in CI settings.
How we do it
We build pipelines in GitHub Actions, GitLab CI or Azure DevOps with OIDC, required reviewers, signed artefacts and automatic rollback, and bring each environment under Terraform or OpenTofu so a new one is a pull request.
What you get
- Pipelines with approvals and rollback
- Terraform or OpenTofu modules and state set-up
- Secrets moved to a managed store
- Runbook for releases and emergency changes
Kubernetes consulting
The problem
EKS, AKS or GKE clusters are several minor versions behind, add-ons are stale, and the bill includes extended-support charges.
How we do it
We scan for removed APIs, upgrade add-ons, rehearse on staging and step versions in place or rebuild blue-green, then leave a calendar and a script so the next upgrade is routine.
What you get
- Upgrade plan per cluster
- Rehearsed upgrade with rollback
- Autoscaling and node-pool review
- Upgrade calendar for the next year
Retained offshore DevOps team
The problem
You need platform capacity every week but cannot justify, or cannot find, a full-time hire in Singapore.
How we do it
A named lead engineer plus specialists as needed, working your shifted hours, through your tickets and chat, on a monthly block of engineer-days that rolls up into a written report.
What you get
- Named lead and backup engineer
- Agreed shared hours and change windows
- Monthly report of work and risks
- Exit plan and full documentation
Typical Singapore engagements
Illustrative patterns from the kinds of teams that contact us. Your scope follows the review.
| Who | What we build | How it usually works |
|---|---|---|
| Seed to Series B SaaS, one AWS account | Separate accounts for production and non-production, Terraform for both, a GitHub Actions or GitLab pipeline with preview environments | Four to six weeks of build, then a small monthly retainer or a handover to your first platform hire |
| MAS-licensed payment or DPT firm | Deploy approvals by a second person, OIDC instead of stored cloud keys, signed images, change records generated from pull requests, just-in-time admin access | Starts with mapping your existing change process against the TRM Guidelines, then fixes in order of inspection risk |
| Regional e-commerce on EKS | Cluster upgraded off extended support, add-ons current, autoscaling and spot capacity tuned before peak sale periods | Rehearsed on a staging cluster, upgraded one minor version at a time, with a rollback plan for each step |
| Singapore headquarters with engineers in several countries | One golden path for new services: templates, policies and a shared pipeline that each country team uses | Platform built with your staff, documented, then run by them with us on call for the hard cases |
| Company replacing a departed DevOps engineer | Knowledge capture of what that person built, secrets rotated, single points of failure removed | Two to three weeks of discovery and clean-up, then an interim engineer or retained team until a hire starts |
Tools we use with Singapore teams
We work in whatever you already run where it is sound, and suggest changes only where they remove risk or toil.
Terraform and OpenTofu
Modules with remote state, plan output attached to every pull request, and drift checks on a schedule.
GitHub Actions, GitLab CI, Azure DevOps
OIDC to the cloud, environment protection rules, reusable workflows instead of copied YAML.
Argo CD and Helm
GitOps for clusters, promotion by image digest, automatic rollback on failed health checks.
Sigstore cosign and SBOMs
Images signed at build and verified at deploy; dependency and image scanning in the pipeline.
IAM Identity Center, Entra PIM, Session Manager
Just-in-time elevation and logged sessions instead of standing admin rights.
Prometheus, Grafana, OpenTelemetry
Service-level objectives and alerts routed to the people who can act on them.
Tools we use alongside it
- Amazon EKS
- Azure Kubernetes Service
- Google Kubernetes Engine
- AWS Backup
- Azure Monitor
- Cloud Logging
- Trivy
- Karpenter
- External Secrets Operator
Facts, versions and vendor details checked 13 November 2026.
How a Singapore engagement starts
Everything happens remotely, in your accounts, during agreed shared hours.
A free 30-minute call
You describe the platform and the pressure: a customer, an inspection, a cluster deadline. We say whether we fit and what the review would cover.
Read-only review
Named, read-only access to cloud, Git and CI. We map the release path and data locations and agree the shared hours and approvers.
Findings and plan
A ranked list with effort in engineer-days. You choose what to fix first; nothing changes in production without your approver.
Build and hand over
Pipelines, infrastructure code and upgrades delivered in small reviewed changes, with runbooks. Then a handover, or a retained team.
A cloud tool we build and run ourselves
We operate our own products on AWS, Azure and Google Cloud, so the inventory and security checks we hand over are ones we depend on.
Guides and related pages
Guides
Questions about DevOps consulting in Singapore
No. Radiatus is registered in Dover, Delaware and as an MSME in India, and our engineers work remotely from India. We work inside your cloud accounts and Git organisation with named access you control. If you need someone on site in Singapore or a Singapore-incorporated supplier, a local firm is the better choice.
Singapore Time is UTC+8 and India Standard Time is UTC+5:30, so Singapore is 2.5 hours ahead all year; neither country uses daylight saving. With the Indian working day shifted earlier, teams usually share six to eight hours, and late-night Singapore change windows fall in the Indian evening.
Yes, for production data. Workloads, backups and logs can all stay in the Singapore region of AWS, Azure or Google Cloud, and our engineers manage infrastructure and pipelines without needing to download records. On Azure, avoid geo-redundant storage if data must stay in Singapore, because the Southeast Asia region pairs with Hong Kong.
We build the engineering controls that MAS’s Technology Risk Management Guidelines and notices describe: segregation of duties in releases, reviewed and reversible changes, MFA and just-in-time privileged access, and timed recovery tests. Whether using us is outsourcing under your rules, and what notification it needs, is for your compliance team; we supply the due-diligence answers.
No. We explain how the rules translate into platform controls and build those controls. Your data protection officer, compliance team or lawyers decide what applies to you and whether the evidence is sufficient.
Yes, remotely. Most work is upgrades off extended support, add-on clean-up, autoscaling and cost, plus GitOps with Argo CD. On EKS, version 1.34 leaves standard support on 2 December 2026 and 1.31 leaves extended support on 26 November 2026, after which AWS upgrades the control plane automatically.
We scope in engineer-days. As planning ranges, a platform review is about 5 to 8 engineer-days, CI/CD for one service family 10 to 20, and one Kubernetes minor upgrade 2 to 5 per cluster. The review produces a written estimate before any build work starts.
Often, but compare like with like. A foreign hire needs an Employment Pass with a qualifying salary floor of S$5,600 a month outside financial services, rising to S$6,000 for new applications from 1 January 2027, and one person cannot cover leave or on-call alone. A retained team trades on-site presence for breadth and continuity.
Running a Singapore platform that needs a DevOps team?
Tell us your cloud and region, how releases happen today, the Kubernetes versions you run and whether MAS rules apply. A senior engineer will reply with what we would look at first and roughly how many engineer-days it takes.






Compliance-first delivery