Browse categories
Explore
Fiverr Pro
English
$
USD
I will create an ai model for you with fine tuning
Dylan S
About this gig
GET YOUR OWN AI MODEL trained on YOUR data, running on YOUR hardware. No API bills. Ever.
WHY A LOCAL MODEL (vs OpenAI fine-tuning)?
You OWN it a real file, yours forever
Zero cost per message no API bills, no subscriptions
Runs offline your data never leaves your machine
Works on a laptop, a server, a small VPS, even offline devices
YOUR DATASET OR DON'T.
Already have a dataset (Q&A pairs, JSON/CSV)? Perfect, we start immediately.
Only have documents (PDF, DOCX, TXT, EPUB)? Add the "Dataset Creation" extra
and I turn them into a quality training dataset.
HOW IT WORKS
1. You send me your data + your goal (tone, behavior, language)
2. I review it honestly if it's not enough, I tell you BEFORE training
3. I fine-tune with QLoRA + Q4 quantization (2-3 epochs, optimized on your dataset)
4. You receive: 2 GGUF files (Q4_0 light + Q8_0 high quality), a model card,
recommended settings, and a test report with real Q&A from YOUR domain
HONEST NOTES (because I test everything I ship):
Quality depends on your data. Good data in = good model out.
You approve the base model I choose (always open-license, commercial use).
Message me BEFORE ordering send 2 lines about your project.
Get to know Dylan S
Dylan S
Chief Marketing Officer General Manager
- FromItaly
- Member sinceSep 2026
Languages
English, Italian, Spanish
I am an experienced CMO and General Manager with a strong background in luxury hospitality and tourism technology. I have founded a SaaS platform for the tourism industry and managed marketing strategies for 5-star luxury resorts. My expertise includes business organization, digital media management, and international guest relations.
My Portfolio
FAQ
What data do I need to provide?
Either a dataset (Q&A pairs in CSV/JSON) or just your documents (PDF, DOCX, TXT, EPUB) with the "Dataset Creation" extra. Minimum ~300 examples for decent results — I'll confirm before starting.
What is GGUF / Q4?
The file format for running local AI models. Q4 = 4-bit quantization: ~75% smaller, runs on normal hardware, minimal quality loss.
Where can I run the final model?
Anywhere: Ollama, LM Studio, llama.cpp on Windows/Mac/Linux, your server, offline. The 1B package runs on almost anything.
Can it handle long conversations?
Depends on the model and the machine you will run it. The model runs with long contexts (2048, 4096+) in Ollama/LM Studio. Training examples are shorter by design — runtime context is not limited.
Why not just fine-tune via OpenAI API?
With OpenAI you rent a model: you pay per use, forever, and your data goes to their cloud. Here you get a file you own 100% — zero recurring costs.
Which base model do you use?
The best open-license one for your task (commercial use) or any alternative. I propose it, you approve before training.
Do I own the final model?
Yes, 100% — file, weights, everything. Full commercial rights (based on the model)
What counts as a revision?
A retraining pass to adjust tone or behavior using the SAME dataset you provided. New data, new languages or new features = new work, quoted separately.
