Vietnam-based AI and data delivery partner
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.
- AI Agent Development & Software Architecture: LangGraph, CrewAI, OpenAI Agents SDK, MCP, LangChain, LangSmith, n8n, Claude (Anthropic), Google Gemini, Mistral, DeepSeek, Ollama, vLLM, AWS Bedrock, pgvector, Milvus, Neo4j, Redis, FastAPI
- Data Engineering & Pipeline Development: Airflow, dbt, BigQuery, Snowflake, Kafka, PySpark, Databricks, Prefect, Airbyte, Pub/Sub, Python, pandas, PostgreSQL, Redshift, Amazon S3, Parquet, Docker, Terraform, Grafana
- Agentic BI Reporting: LangGraph, dbt Metrics, MetricFlow, BigQuery, Snowflake, MCP, Claude, OpenAI GPT-4o, Cube, LookML, Power BI, Tableau, Metabase, Plotly Dash, Streamlit, Jupyter, Slack, Teams, Google Sheets, Excel
- Web Scraping & Automation: Playwright, Selenium, Selenium Grid, Python, BigQuery, Scrapy, Headless Chromium, pandas, Pydantic, JSON Schema, PostgreSQL, Snowflake, Amazon S3, Airflow, Celery, Redis, Docker, Kubernetes, Grafana, Sentry
- Cloud Architecture & Cost Optimization: AWS, Google Cloud, Azure, Terraform, Docker, Kubernetes, GitHub Actions, Cloud Run, GKE, ECS Fargate, AWS Lambda, Cloud Functions, CloudWatch, Grafana, Prometheus, Datadog, OpenTelemetry, BigQuery, Athena
- System Operations & Maintenance: Grafana, Sentry, LangSmith, GitHub Actions, Prometheus, Datadog, OpenTelemetry, CloudWatch, Terraform, Docker, Kubernetes, Airflow, dbt, Playwright, PostgreSQL, Redis, Next.js, NGINX, Vercel, Claude
Selected clients
Trusted by product teams in 15+ countries
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:
SeaSure
Graspr AI
Zeno
Ahrvo
Workflow preview
Watch the full-service workflow in motion
See how research, build, automate, harden, and handoff can stay inside one senior delivery loop instead of scattering across tools and vendors.
Founder-led delivery loop
Founder-led delivery from workflow audit to production hardening
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.
Tran Tien Van
Founder and Principal EngineerHo Chi Minh City, Vietnam
Tran Tien Van
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.
- LangGraph multi-agent systems
- Data pipeline engineering
- Agentic BI and reporting
- Web scraping at scale
- Evaluations and guardrails
- Upwork rating from 71 reviews
- 5.0
- Projects delivered
- 150+
- Countries served
- 15+
- Years of development experience
- 5+
Service lanes
Services built around the bottleneck
Choose the lane that matches the workflow, reporting, automation, or data problem your team needs to scope, ship, and harden.
Autonomous execution
AI Agent Development & Software Architecture
Design AI agents that reason, call tools, and complete production workflows with clear guardrails and human review.
Multi-step agent workflows for support, research, and operationsTool-connected execution across APIs, documents, and internal systemsHuman review, audit trails, and fallback handling for sensitive actionsCrewAILangGraphOpen service page
OpenAI
ClaudeReliable data movement
Data Pipelines
Build resilient batch and streaming pipelines that move data cleanly from source systems into warehouses, reports, and AI-ready products.
Batch and streaming ingestion designed around real freshness needsWarehouse-ready transformation models for reporting and AI consumersMonitoring, alerting, and replay-aware operational controlsAirflowOpen service page
KafkaBigQuerydbtInsight on autopilot
Agentic BI & Reporting
Build governed LangGraph reporting agents that query your semantic layer, validate SQL, explain KPI movement, and deliver Slack-native narratives.
Six-agent LangGraph workflows for planner, SQL, validation, execution, visualization, and narrationBigQuery, Snowflake, or Postgres integration through MCP with auditable tool calls80-150 question eval suites and validation gates for CFO-grade reportingLangGraphMCPBigQuerySnowflakeOpen service pageHard-to-get data
Web Scraping & Automation
Collect structured data from dynamic and protected sources with browser automation, proxy strategy, and production-grade delivery.
Playwright and Selenium workflows for dynamic and protected targetsProxy-aware routing, retry orchestration, and anti-detection handlingStructured parsing and validation for usable downstream dataOpen service page
Playwright
SeleniumScrapyProxyScale without waste
Cloud Cost Optimization
Reduce cloud waste while protecting delivery speed through workload analysis, right-sizing, scheduling, and targeted re-architecture.
Spend visibility tied to workloads, teams, and business pressureRight-sizing, scheduling, and storage changes that create durable savingsArchitecture recommendations that protect reliability while lowering wasteOpen service page
AWS
GCP
Azure
DockerLegacy to leverage
Platform Modernization
Modernize brittle data and integration stacks into platforms that support reporting, automation, and AI delivery with less operational drag.
Warehouse and integration contracts that are easier to extend safelyMigration sequencing that protects live reporting and operationsDeployment, rollback, and support patterns teams can sustainOpen service page
FastAPI
PostgreSQLS3Cloud RunAfter launch
System Operations & Maintenance
We keep AI agents, data pipelines, scrapers, apps and websites running after launch: monitored, maintained and reported every month.
Monitoring and alerts on every system we runIncident response with a written cause and fixUpgrades ahead of model, API and runtime deadlinesGrafanaSentryLangSmithGitHubOpen service page
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"
On camera with clients
Face to face with the teams we build for
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 clientsSelected work
Systems we've shipped into production
Our clients' products, live today: an AI agent platform, learning, cybersecurity, marine, fitness and an AI mentor. Each has a full case study.
goalice.ai
Client projectLiveSmall business operations
A supervisor agent that runs 70+ business tools for small-business owners
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
graspr.ai
Client projectLiveEducation
RAG tutoring that lifted answer accuracy from about 80% to 95%+
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%
seasure.ai
Client projectLiveMarine
One LangGraph agent for every boat subsystem
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
yojin.io
Client projectLiveCybersecurity and compliance
A fine-tuned Mistral 7B behind an EU-sovereign compliance platform
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
AI training programs that reach users only above 85% confidence
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
consequentialflowers.ai
Consequential FlowersClient projectLive
Education and mentoring
An AI mentor built from each partner's own content within 24 hours
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
Data and BI projects
Pipelines and datasets we delivered, each shown with the dashboard built from its data.
NEM wholesale electricity: dispatch monitor
Energy markets
Real-Time Electricity Market Pipeline on AWS
Energy market analytics teamClient project
5 min
ingestion cadence
Twitter/X: follower harvesting and concentration
Social intelligence
Twitter/X Follower Monitoring at Scale
Social intelligence clientClient project
30,000+
profiles tracked per daily run
LinkedIn employee advocacy: ESG signal extraction
ESG research
LinkedIn ESG Research Panel
Quantitative research teamClient project
1.4M
employee posts scored for E, S and G
SSRN research corpus: attention and authorship
Academic research
SSRN Research Papers and Authors Dataset
Academic research clientClient project
239,037
papers delivered, 43 fields each
Engagement models
Ways to work with Van Data Team
Choose the engagement shape that fits your team, then start with a free call.
- 01
Strategy Sprint
Scope and de-risk
Validate the workflow before a bigger build.
Typical outputs
- Discovery
- Workflow audit
- Architecture direction
- Quick-win backlog
- Delivery estimate
- 02Recommended
Production Build
Ship the system
Best for business-ready AI, automation, or data delivery.
Typical delivery
- End-to-end delivery
- Testing and deployment
- Docs and handoff
- 30-day optimization
- Weekly reporting
- 03
Embedded Partner
Senior support
Ongoing support inside complex AI or data roadmaps.
Typical support
- Priority response
- Architecture support
- Performance tuning
- Ops improvements
- Flexible sprint execution
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.
FAQ
Questions teams ask before they reach out
Straight answers on what Van Data Team builds, who it fits, and how projects start.
What does Van Data Team build?
Van Data Team builds AI agents, reporting workflows, data pipelines, automation systems, and modernization work for teams that need production-ready delivery.
Who is the best fit for Van Data Team?
The strongest fit is a team with a real workflow bottleneck, a delivery owner on the client side, and a need for senior implementation rather than slideware.
Can you work inside an existing stack?
Yes. Many engagements start from an existing stack that needs stabilization, better observability, or a new AI and data layer added on top.
How do projects usually start?
Projects usually start with a scoped conversation, a workflow review, or a short strategy sprint that clarifies the bottleneck, risks, and recommended next step.
Why work with a Vietnam-based delivery partner?
A Vietnam-based delivery model can keep overhead lean while still giving global teams senior ownership, strong communication, and timezone overlap that supports steady delivery.
How is a project priced?
Each engagement gets a fixed price in a written scope after a free 30-minute call: a short Strategy Sprint to scope and de-risk the work, a fixed-quote Production Build, or ongoing Embedded Partner support billed monthly or by the hour. Each service page lists its typical range.
Who owns the code and the data?
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.
How do we work together day to day?
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.
What happens after handoff?
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 under an Embedded Partner engagement.
Free strategy session
Get a 30-minute AI and data consultation
Bring the current process, blockers, and decision pressure. The output is a practical next-step plan, not a vague sales call.
you@company.com