I will build a custom machine learning mvp

Yes. For prototypes, I often start with open datasets (Kaggle, HuggingFace, public APIs) that match your use case, so you can validate the concept before investing in data collection. Once it works, we plan the data strategy for v2. Message me with the problem you're solving.

This gig builds an ML prototype or MVP from scratch — model, demo, API. The fine-tuning gig is specifically for adapting large open-source LLMs (Llama, Mistral) to your domain with LoRA/QLoRA. If you're not sure which fits, message me your use case and I'll point you to the right one.

Yes, that's the point. I build prototypes the same way I build production systems — clean code, real evaluation, Docker-ready. The Premium tier is already production-grade. For Basic/Standard, scaling later means adding monitoring, retraining pipelines, and load handling, not rewriting.

No problem. I work across both — XGBoost and scikit-learn where they fit, PyTorch and HuggingFace transformers where they don't. I'll recommend the simplest approach that solves your problem. Throwing deep learning at every problem is how prototypes get expensive and slow.

Yes, happy to sign one before you share anything sensitive. Just send it over before we kick off. For most prototypes the standard Fiverr terms cover confidentiality, but a custom NDA is fine if your legal team requires it.

Depends on the problem. Classical ML can work with a few hundred clean examples; deep learning usually wants thousands. If you're under that, we can use transfer learning, data augmentation, or synthetic data. Message me what you have and I'll tell you honestly if it's enough.