Browse categories
Explore
Fiverr Pro
English
$
USD
I will parse your emails into clean structured data
Kaze
About this gig
Your team copies order confirmations, invoices or booking emails into a spreadsheet by hand. I turn that inbox into a reliable data feed.
I spent two years building and hardening an email-to-database pipeline that runs in production every night: IMAP and Gmail connections, sender detection, one parser per template with an LLM fallback, forwarded-email unwrapping across providers and languages, thread and duplicate handling, currency normalisation, and a correction loop that learns from every manual fix.
WHAT YOU GET
- A parser per email type, with a documented JSON schema
- Real fixtures and automated tests, so accuracy is measured, not promised
- No silent failures: anything unparsed lands in a review queue
- Handover docs so your own developers can add the next sender
Any language, any provider. TypeScript, Node or Python, OpenAI or Claude, and your database, Google Sheets or a webhook as the destination.
Send me 3 to 5 sample emails (redacted is fine) and I will tell you honestly what is parseable and what is not, before you order.
Get to know Kaze
Kaze
I automate what slows you down
- FromUnited States
- Member sinceOct 2022
- Avg. response time1 hour
Languages
French, English
Developer and automation specialist with experience building production-ready tools across logistics, e-commerce, and SaaS.
My expertise: automation & bots, full-stack web dev (React, Next.js, Node.js), data dashboards, Gmail automation, and API integrations.
I've built and operated my own SaaS platforms — so I deliver code that works in production, not just in theory. Fast delivery, clean code, no unnecessary complexity.
Other Software Development Services I Offer
FAQ
Do I have to give you access to my inbox?
No. I develop against sample emails you export and redact. If you want the live connection wired, use an app-specific password or a dedicated forwarding address you can revoke at any time.
What accuracy can I expect?
On stable templates, near 100% — those are parsed by rules, not by a model. On free-form emails the LLM fallback usually lands 90-95%, and anything below the confidence threshold goes to a review queue instead of your database.
Can you parse emails in languages other than English?
Yes. The pipeline I run in production handles English, French, German, Spanish and Italian senders, including forwarded messages where the original sender is buried in the body.
Can you deliver this in Zapier, Make or n8n?
No, and I would rather say so upfront. I write custom code you own. No-code tools are cheaper to start and fine for simple flows, but they struggle with template drift, duplicates and retries, which is most of the work here.
