Senior engineers take agent workflows, data pipelines and reporting from prototype to production in weeks, with evals, guardrails and monitoring built in. No bloated agency layer: the founder leads every engagement.
Selected clients
Startups and established companies rely on Van Data Team for AI agents, data platforms and the software around them. A few of the products we have helped build:
Consequential FlowersAhrvoEllipsis EarthVoodooStar AnalyticsThe problems we solve for startups and growing businesses, the AI agents and data systems we build, how founder-led delivery works and how to book a free 30-minute call.
Our clients' products, live today: an AI agent platform, learning, cybersecurity, marine, fitness and an AI mentor. Each has a full case study.
Client projectLiveSmall business operations
We built the multi-agent core of alice/ai: a LangGraph Supervisor that routes every request to specialised sub-agents, with retrieval over each customer's data and tracing in LangSmith, on FastAPI and AWS ECS Fargate.
70+
MCP tool integrations under one supervisor
Client projectLiveEducation
We built Graspr's retrieval layer: pgvector embeddings and semantic search with LangChain and Claude, behind the tutors, quizzes and flashcards it generates from students' own PDFs and lectures.
95%+
RAG answer accuracy, from about 80%
Client projectLiveMarine
A FastAPI backend of eight domain modules that routes each question to the agent for that subsystem, across Claude 3 and GPT-4.
8
domain modules in production
Client projectLiveCybersecurity and compliance
We fine-tuned Mistral 7B for security and compliance and paired it with retrieval over a Neo4j threat-and-compliance graph and ChromaDB, orchestrated in LangGraph, so each organisation gets a prioritised action plan.
424k
fine-tuning samples, plus 106k for validation
Client projectLiveHealth and fitness
Multi-provider routing across Claude 3 Haiku, OpenAI and Gemini for program design, exercise generation and calorie tracking.
85%
confidence gate on every generated program
Education and mentoring
PRIM, a white-label AI mentor companion, and SeedVision, which turns a content library into AI-ready "Seeds". Claude and GPT-4 with LangChain on GCP.
~5 s
response time on a 200K-token context
Pipelines and datasets we delivered, each shown with the dashboard built from its data.
Energy markets
Energy market analytics teamClient project
5 min
ingestion cadence
Social intelligence
Social intelligence clientClient project
30,000+
profiles tracked per daily run
ESG research
Quantitative research teamClient project
1.4M
employee posts scored for E, S and G
Academic research
Academic research clientClient project
239,037
papers delivered, 43 fields each
Tran Tien Van leads the work from the first workflow audit through architecture, implementation, and hardening, so business context and technical decisions stay aligned all the way into production.
Founder and Principal EngineerHo Chi Minh City, Vietnam
Founder-led AI agent and data engineering delivery from Ho Chi Minh City, Vietnam.
The person you talk to is the person accountable for the architecture and the quality of what ships.
The same engineer handles the workflow audit, architecture direction, implementation, and hardening, so decisions and execution stay in one loop.
Verified review signal
5.0 · 71 reviews on Upwork
Chris K.
5.00 / 5
FT Platform Phase #2
"Great backend developer, highly recommend!"
Gilad B.
5.00 / 5
Phase 0: Design a granular data schema and structure, and full tool flow
"Very knowledgeable and professional. Good communication"
Julio D.
5.00 / 5
Web scraper in R
"Tran was great, very knowledgeable and quick responses"
Scoping calls, demos, reviews and handoffs, run by the founder on camera and in your working day.
Client names, captions and project details are hidden.
How we work with clientsChoose the engagement shape that fits your team, then start with a free call.
Scope and de-risk
Validate the workflow before a bigger build.
Typical outputs
Ship the system
Best for business-ready AI, automation, or data delivery.
Typical delivery
Senior support
Ongoing support inside complex AI or data roadmaps.
Typical support
Not sure which one fits?
Start with a free 30-minute call with the founder. We recommend the shape that fits your team, even when that is a smaller one.
Built with
We choose the cloud, models and tools for each system instead of one vendor for everything. These are the ones we ship with most.
Trademarks belong to their owners. Listing a platform does not mean a partnership with or an endorsement by its owner.
Straight answers on what Van Data Team builds, who it fits, and how projects start.
Van Data Team builds AI agents, data pipelines, agentic BI reporting, web scraping and automation, and cloud architecture with cost optimization, and operates and maintains these systems in production. It is a founder-led AI and data engineering practice in Ho Chi Minh City, Vietnam, working with clients worldwide.
Product and operations teams with a real workflow problem, someone on their side who owns the outcome, and a need for a system running in production rather than a slide deck or a demo.
Yes. Most projects start from what a team already runs, such as a data warehouse, a CRM, internal APIs or cloud accounts, and add an AI or data layer to it or make it more reliable. We extend your stack rather than replace it, unless replacing a part costs less.
With a free 30-minute consultation or a short project brief. We look at the workflow, the systems involved and your constraints, then send a written scope with a fixed quote before any work starts.
Senior engineering on a lean cost base, and working hours that cover the APAC day, the morning in the UK and Europe and the start or end of the US day. At Van Data Team the founder who scopes the work also builds it, in English, with progress visible in your own tools.
Each project gets a fixed quote in a written scope after a free 30-minute call, based on the systems to connect, the data volume and how much has to run unattended. There is no published rate card, and support after launch is quoted the same way, as a fixed monthly scope.
You do. Once the invoice is paid in full, everything written for you is yours, and cloud, API and AI model accounts are opened in your name. Your data stays confidential during the project and for three years after it, and a data processing agreement is signed before any personal data is shared.
In English, in writing and on calls, on Vietnam time (UTC+7), which covers the APAC working day, the morning in the UK and Europe, and the start or end of the day in the US. You work directly with the founder who scopes and builds the system, and progress is visible in your own repository and tools.
Every system ships with documentation and a handoff. If a deliverable does not meet the agreed scope and you report it within 30 days of acceptance, it is fixed free of charge, and monitoring and improvements can continue on a fixed monthly scope.
Free strategy session
Bring the current process, blockers, and decision pressure. The output is a practical next-step plan, not a vague sales call.
