50+ specialists. Three AI disciplines. One stack.
We're an enterprise AI implementation partner — engineers who ship production systems, product specialists who scope what gets built, and a growth team that lands the right engagements. Built on a modern AI-native stack tuned for shipping, not slides.
How 50+ people split across the work.
Engineering-heavy by design. Most of an enterprise AI engagement is software work — integrations, data plumbing, evaluation pipelines, ops. Products keeps scope honest; Growth keeps the engagement pipeline healthy.
AI Engineering
Ship the production systems — RAG pipelines, agent orchestration, integrations into CRM/ERP/helpdesk, evaluation pipelines, ops. Core craft is software engineering, sharpened on the AI stack.
- Senior AI Engineers
- Backend Engineers (Java / Python)
- Frontend Engineers (React / React Native)
- Data & ML Engineers
- DevOps / Cloud Engineers
AI Products
Translate business outcomes into scoped engagements. Workflow inventories, use-case ranking, solution architecture, evaluation design, UX of AI features that real users will actually adopt.
- Solution Architects
- Product Managers
- Product Designers
- AI Evaluation Specialists
- Engagement Leads
AI Growth
How clients find us, how engagements land, how relationships deepen post-launch. Sales, marketing, customer success, partnerships across SEA and the wider international market.
- Enterprise Sales
- Customer Success Managers
- Marketing & Content
- Partnerships
Modern AI-native, pragmatically enterprise.
Frontier LLMs where they earn their cost; battle-tested infrastructure underneath. Java and Python on the back end, React + React Native on the front. We pick per use case — no fixed allegiance to any vendor or framework.
AI & language models
Frontier LLMs and the production glue around them.
- Anthropic Claude (Opus / Sonnet)
- OpenAI GPT family
- Google Gemini
- Mistral & open-weights (Llama, Qwen)
- PhoBERT, ViT5 (Vietnamese-tuned)
- BGE-M3, mE5 (multilingual embeddings)
- LangChain · LlamaIndex
- Custom agent orchestration
- PyTorch · Hugging Face
Retrieval, knowledge & data
Vector + graph + relational, picked per use case.
- PostgreSQL + pgvector
- Qdrant · Weaviate · Pinecone
- Elasticsearch / OpenSearch
- Neo4j (knowledge graphs)
- Redis (cache, rate limits)
- Kafka · RabbitMQ (event streams)
- dbt · Airflow (data pipelines)
Backend & APIs
Performant services in the languages your team already runs.
- Python — FastAPI · Django · Flask
- Java — Spring Boot · Spring Cloud
- Node.js / TypeScript — Next.js · NestJS · Express
- gRPC · REST · GraphQL
- OAuth2 / SAML / OIDC
Frontend & mobile
Production UIs that AI features can plug into cleanly.
- React · Next.js 15 (App Router)
- React Native (iOS + Android)
- TypeScript end-to-end
- Tailwind CSS · shadcn/ui
- Radix · Framer Motion
- Streaming UIs (SSE · WebSockets)
Cloud, infra & DevOps
Multi-cloud by default. Customer chooses the region.
- AWS · Azure · GCP
- Docker · Kubernetes
- GitLab CI · GitHub Actions
- Terraform · Pulumi
- Grafana · Prometheus · Sentry
- Plausible (cookieless analytics)
Enterprise integration
Where AI lands in your existing systems.
- Salesforce · HubSpot · Pipedrive
- SAP · Oracle · NetSuite
- Zendesk · Intercom · Freshdesk
- Microsoft 365 · Google Workspace
- Webhooks · ETL · iPaaS
AI-native development practices.
We don't just build AI products — we build them with AI. Every engineer pairs with a coding agent every day. The result: faster iteration, fewer regressions, more time for the parts that need human judgment.
Vibe coding
AI-native pair programming. Engineers describe intent in natural language; the AI proposes code, tests, and refactors; the engineer reviews, accepts, and ships. Coined by Andrej Karpathy in early 2025, now standard practice across our team — faster iteration without losing the engineering rigor.
AI-assisted code review
Every merge request gets an AI review pass before a human reviewer touches it — catches the obvious smells, security pitfalls, and missing tests, so human reviewers focus on architecture and intent.
Evaluation-first development
AI features ship with eval pipelines from day one — gold-standard test sets, automatic regression checks on every PR. Retrieval drift, hallucination rate, and citation accuracy are tracked the same way we track latency.
Voice & multilingual by default
Vietnamese + English fluency, code-switching, voice channels (TTS / STT), and dialect awareness baked into the engineering process — not retro-fitted at the end.
Want this team on your engagement?
Start with an AI Audit. We'll map your highest-leverage use cases and propose the right cross-functional pod for the work.