Browse categories
Explore
Fiverr Pro
English
$
USD
I will extract structured data from text with an ai llm into validated json
About this Gig
Customer messages, emails, listings or notes arrive as free text and someone copies the details into a sheet? I build an extractor that sends each text to an LLM and returns a record in a fixed JSON format, then checks it before it reaches your system.
What you get: a Python script with a strict schema for your fields, validation of every record, a repair step for fixable output, a check that quoted evidence really appears in the source text, and a test report on your sample texts showing where it gets things wrong.
Why me: I built this for my own AI project (synthetic data, not a client deployment): 12-field records from customer messages with strict JSON schema output, validation, evidence checks and an evaluation gate in CI. I report real misses, including a live run that fell below my own accuracy floor, instead of hiding them.
Not included: OCR of scans, model fine-tuning, hosting. You pay your own LLM API costs. Works with OpenAI-style APIs.
Technology:
Excel
•
Python
My Portfolio
FAQ
What do you need from me?
5 to 25 real example texts, the fields you want, and an LLM API key for testing (or I use my own and you reimburse the cost). Remove personal data first.
How accurate will it be?
I do not promise a number. You get a test report on your sample texts showing which fields it gets right and where it fails, so you can decide.
Do you handle scanned documents or images?
No. This is for text that already exists as text. Scans need OCR, which I do not offer here.
Which models and providers?
OpenAI-style APIs, and models you serve behind an HTTP endpoint. You pay the API costs. Hosting is not included.
