VU Solutions · Software & AI engineering · Vietnam

I build software products, and the AI agents that help run them.

I'm Vu, a software developer and the founder of VU Solutions. I've shipped .NET and React products for fintech, pharma, racing and childcare clients since 2019. These days I also design agentic AI workflows, and I use them to run my own work.

Vo Hoan Vu
Vo Hoan Vu · Founder

01 · Right now

What I'm working on

VU Solutions is my registered single-member company. Client work, a startup and my own product all run through it.

Client work · Remote

Developer for OWNA

I work as a remote developer for OWNA, an Australian SaaS company serving the childcare sector, contracting through VU Solutions.

Startup · CTO

CTO of TKNQ

TKNQ is an early-stage Australian startup building all-in-one software for Brazilian Jiu-Jitsu gyms. I own the technical side: architecture, AI research and delivery.

My product · Live demo

CodeJitsu

An AI coach that watches BJJ sparring footage and turns it into a timestamped breakdown, a grade and the drills that fix what went wrong. It's also TKNQ's R&D testbed.

02 · Services

What I can do for you

Product engineering

Web products from the first prototype to production: ASP.NET Core and C#, React and TypeScript, PostgreSQL, Google Cloud and Cloudflare. I take features end to end, from database migration to deployed and monitored.

Agentic AI workflows

I design, orchestrate and evaluate multi-step AI pipelines. That means splitting a job into focused agents, adding verification passes, capping cost and keeping a person in the loop.

I run my own development team of AI agents, and my company's bookkeeping and tax workflow runs on MISA with AI assistance. I've seen where agents break, not just where they shine.

Technical leadership

For an early-stage team that needs a CTO's judgment before it can hire one: architecture, delivery plans, cloud operations and cost control. That's my role at TKNQ today.

03 · How I work

Straight answers, both ways

Tell me what went wrong

I'll tell you plainly when an idea, an estimate or a line of code is wrong, including my own. I ask for the same in return. If something I shipped missed the mark, say so directly. Problems found early are cheap; problems softened into politeness are expensive.

Principles I build by

  • Small steps you can verify. I break work into pieces that can each be checked by a test, a review pass or a person.
  • AI only where it earns its place. I've replaced an AI feature with a plain API call when that was faster, cheaper and more reliable.
  • Cost and failure are part of the feature. Quotas, refunds and retries belong in version one, not after the first surprise bill.

04 · Selected work

Products I built and run myself

CodeJitsu landing page: "AI that reads a BJJ sparring tape"

CodeJitsu: an AI coach for BJJ sparring footage

Upload a sparring video and Gemini returns a timestamped breakdown, a grade and the drills to work on, then drafts the next class plan. Built with a team of five AI agents.

Semantic search · Embeddings

JRE Clipper: "What would Joe Rogan say?"

Ask about any topic and get the exact moments it came up across the Joe Rogan Experience archive. Transcripts are turned into embeddings, so search works by meaning, and every result jumps straight to the moment it was said.

  • Data embeddings
  • Semantic search
  • Vertex AI

05 · Contact

Tell me what you're building

A product to build, an AI workflow that isn't working, or a team that needs technical leadership. Every message comes straight to me.