Tai Huynh
Freelance · AI & data tooling · open to commissions

AItoolsthatshipcleandatatoyourbusiness.

I’m Tai. I build production AI tools — data pipelines, scrapers, and agent infrastructure — for businesses that need the open web to behave like a typed API. The page below is one of them. It’s live, not a screenshot.

Try the demoHire mereplies within 24h

The demo, live.

deployed on vercel · gemini 2.5 flash · ~$0.0008 / scrape

Paste any public URL and describe the fields in plain English. Get a typed table, CSV, and JSON in seconds. Try one of the pre-loaded specimens or your own page.

Live demo—:—:— UTC

10 requests · ip · day  ·  public pages only  · full pipeline →

What I build for clients

Three shapes of work.

These are the patterns I’ve shipped most recently. Your project probably looks like one of these or a hybrid — drop me a line and we can scope it.

01

Competitor price & inventory monitoring

Daily scrapes of a competitor catalog, diffed over time, with alerts on price moves or stock-outs. Output is a typed JSON feed or a CSV you can drop into BI.

Best for: e-commerce, marketplaces, hospitality, anywhere price-moves matter.

02

Lead enrichment from public sources

Turn a list of company URLs into a structured row per company — about-page summary, public contact, tech stack hints, employee count from public sources.

Best for: sales/BD teams who buy lists and want them deduped, enriched, and ready for outreach.

03

Structured-data pipelines for AI agents

Wrap any public source as a typed endpoint your agent can call. Schema-validated, rate-limited, robots-respecting, with the audit trail your compliance team will ask for.

Best for: teams building RAG / agent systems that need clean, reliable, non-PII web data.

How I work

No tricks. No black hats.

Security-first

Every URL passes an SSRF guard before a packet leaves the process — cloud-metadata addresses, private ranges, non-HTTP schemes all rejected. robots.txt enforced, not just consulted.

Honest about gaps

I tell you what the build does and does not cover. The how-it-works page calls out the DNS-rebinding gap on this demo and what the full client build adds — no hidden surprises.

Production discipline

Typecheck, tests, structured errors, cost ceilings, redacted logs. The demo is deployed the same way I would deploy your build — no shortcuts that the "real" version would clean up later.

Senior judgement

I push back early when a design will fail in three months. Most of the value I bring is not the code I write but the code I talk you out of writing.

Contact

Want one of these for your business?

The fastest path is an email with: what you’re trying to extract, from where, and how often. I’ll come back within 24 hours with a scope and a price — or a clear “no, here’s why” if it’s outside my wheelhouse.