NKKTech doesn't pick favorites. We're cloud-agnostic, model-agnostic, and stack-agnostic — we route each project to the right partner based on the client's existing infrastructure, compliance requirements, residency rules, and economics. Here's the full partner ecosystem we operate in for production AI engineering across our 120+ projects since 2018.
Where our clients' AI workloads physically run. We default to AWS for North American and ASEAN clients, Google Cloud for ML-heavy workloads, and Azure for HIPAA-regulated US healthcare and EU residency-sensitive deployments. All three are first-class in our stack.
Primary cloud for ~60% of NKKTech production deployments. We use Bedrock for managed LLM access (with BAA for HIPAA-touching workloads), Lambda for serverless AI orchestration, ECS/EKS for containerized agent workloads, S3 + RDS Aurora for storage. We have engineers with active AWS Solutions Architect and AWS ML Specialty certifications.
Default for ML-research-heavy workloads and Japan-based clients with existing Google relationships. We use Vertex AI for fine-tuning and model serving, BigQuery for analytics-side AI workloads, Cloud Run for stateless agent deployments. Strong choice when clients need Gemini's long-context capabilities.
Default for HIPAA-regulated US healthcare and EU residency-sensitive workloads. Azure OpenAI Service in West Europe and Sweden Central for EU residency, with BAA available for HIPAA. We also use Azure Cognitive Services for document intelligence in regulated workflows.
Cost-conscious option for smaller production workloads (~$1M ARR clients). Where AWS/GCP overhead doesn't pay back, DO Droplets + Managed PostgreSQL + Spaces give 60-70% cost reduction at acceptable operational tradeoff.
The LLMs and embedding models that power AI features we ship. We're model-agnostic and route per task to the right provider; we maintain operational expertise across all five.
Default for general-purpose reasoning, agents, and tool calling. GPT-4o for flagship workloads, GPT-4o-mini for high-volume sub-tasks. We hold a Tier 4-5 organization account with zero-retention enabled. Not used for HIPAA (no BAA available for standard accounts).
Claude 3.5 Sonnet for long-context tasks (legal, medical document review) and customer-facing applications where Claude's safer tone is preferable. Available directly and via AWS Bedrock + Google Vertex; we route based on residency requirements.
Gemini 1.5 Pro for workloads requiring 1M+ token context windows — multi-document analysis, large codebase understanding, long meeting transcripts. Available via Vertex AI for regional residency.
Mistral Large 2 for European data residency requirements (Mistral is EU-headquartered) and as a quality alternative when clients want non-US LLM dependencies. Also a strong choice for self-hosting via the open-weight models.
Voyage embed-3-large for top-quality English embeddings; Cohere multilingual-v3 for multilingual retrieval (English + Japanese + Korean + Chinese + Vietnamese in the same index). Cohere Rerank 3 is our default reranker in production RAG.
Enterprise platforms we routinely integrate with on behalf of clients. Not vendor relationships in a marketing sense — operational expertise from real client work.
Oracle Database, Fusion ERP, and OCI for enterprise client integrations. Common in fintech and manufacturing clients with existing Oracle stacks. We've delivered Oracle-to-modern-stack migrations and Oracle-integrated AI workflows.
Watson AI services for enterprise NLP workloads, IBM Cloud for clients with existing IBM relationships. Less common than AWS/GCP/Azure but operationally familiar to our senior team.
Japanese enterprise IT services partner — many of our Japanese clients run alongside or in coordination with NTT Data's broader engagements. We've collaborated on multi-vendor projects where NKKTech handles the AI layer and NTT handles the broader IT modernization.
Billing, subscription, and payment infrastructure for SaaS clients. Stripe Atlas for our own Singapore Pte. Ltd. entity. We've shipped Stripe billing integration as part of est-invoice and several SaaS client projects.
Headless CMS for content-heavy AI products and enterprise documentation. Our own nkktech.com is built on Sanity. Strong choice when AI-generated content needs editorial review workflows.
Frontend deployment platform for Next.js applications. Default for SaaS and marketing-site frontends. nkktech.com itself runs on similar self-hosted Next.js standalone infrastructure.
Founder Tony Nguyen spent 10 years building enterprise software inside Japanese companies before founding NKKTech in 2018. The relationships and operational understanding of how Japanese enterprises work — from contract structure to communication patterns to release discipline — are a structural advantage for our Japan-targeted client work.
Tony's prior engineering work on Toyota systems informs our automotive IoT and manufacturing-systems projects. We've delivered AI workflows for Tier 1 and Tier 2 Japanese automotive suppliers since 2019.
Background in Sony consumer electronics and AI imaging shapes our work for Japanese media, entertainment, and IoT clients. The release discipline expected by Japanese enterprise (every release runs through a formal Quality Gate review) is built into how we ship every project.
Telecommunications and AI investment ecosystem context — useful when our Japanese clients have SoftBank Robotics or SoftBank-funded vendors in their stack. We've integrated AI agents with SoftBank Pepper for retail and museum deployments.
E-commerce platform integrations and Rakuten Pay payment flows for Japanese SaaS clients. Cross-border e-commerce workflows that span Rakuten ecosystem to international markets are a frequent integration pattern.
Formal partner program status with the platforms we deploy on most. Status as of 2026-05. Active engagement-track activities (certifications, completed deployments, reference customers) progress most of these toward higher tiers throughout 2026.
Standard Tier, Advanced Tier in progress
We hold Standard Tier status; certified solutions architects on the team are progressing the engagement toward Advanced Tier qualification. Listed in AWS Partner Solutions Finder under AI/ML Services category for ASEAN and JP regions.
Registered Partner
Registered partner; engineers hold Google Cloud Professional ML Engineer and Professional Cloud Architect certifications. Engagement-track activities for Specialization status in 2026.
Solutions Partner application pending
Application submitted for Solutions Partner designation in Data & AI Azure category. Decision expected H2 2026.
Application planned
OpenAI's solution partner program for enterprise integration vendors. Application timed with OpenAI's annual partner application window.
Registered
Registered as Anthropic Solution Partner for Claude API integration projects. Direct access to enterprise sales and technical support for client engagements.
Registered
Sanity Partner Agency for headless CMS implementation projects. Active reference customer (nkktech.com runs on Sanity).
Independent third-party audits of NKKTech Global JSC (Vietnam HQ) covering quality management and business continuity. Both certifications are renewed annually under surveillance audit and are verifiable directly with the certifying body — neither is a self-declaration.
Issued by: TQC CGLOBAL
Certificate ID: TQC.01.7910
Valid until: 2029-04-03
Quality Management System — Software development services. Annual surveillance audits ensure ongoing compliance. Verifiable at tqc.vn.
Issued by: CGLOBAL Global Inspection & Certification Network
Certificate ID: CGLOBAL.10.0106
Valid until: 2029-04-03
Business Continuity Management System — Software development services. Covers incident response, disaster recovery, and operational resilience. Verifiable at cglobal.us/verify-certificate.
For client engagements, the partner stack we recommend is derived from the client's actual requirements, not from incentive arrangements with any vendor. We don't receive referral commissions or affiliate kickbacks that would bias our architecture recommendations. If a client's use case is better served by Anthropic Claude than by an OpenAI model — even when we'd earn nothing from the recommendation either way — we'll say so and architect around it.
For our own product (est-invoice) and internal operations, we use what works: Anthropic Claude for legal-language tasks, OpenAI for general reasoning, Google Vertex for long-context analysis, AWS for production infrastructure, Sanity for content management, Vercel for frontend deployment. The stack is opinionated but we're always re-evaluating as the landscape shifts.
If you're a technology vendor considering a partnership with NKKTech for distribution into the Vietnam, Singapore, or Japan markets, reach out via the contact form. We're selective about formal partnerships — we'd rather have five deep relationships we can really stand behind than fifty logos we can't.