Frequently asked questions
Straight answers to the questions we hear most often. If yours is not here, contact us and we will answer it directly.
How do you keep my data secure?
Data is transferred over encrypted channels and stored encrypted. Access is limited to the engineers on your project, and we can work entirely inside your own cloud account if you prefer.
Who owns the model weights?
You do. Our agreement assigns ownership of the trained weights to you, subject only to the license terms of the open-weight base model.
Will you use my data to train models for other clients?
No. Your data is used only for your project and is never shared, sold, or reused.
What hardware do I need to run the finished model?
It depends on the model size. Small models can run on a single workstation or modest cloud server; larger ones need one or more GPUs. We size this during scoping and document it in the deployment guide.
How long does a project take?
Discovery takes one to two weeks. Distillation and fine-tuning projects usually take four to ten weeks. Deployment takes two to four weeks. These are estimates that depend on your data and review cycles.
What is the minimum dataset size?
There is no universal minimum. Instruction tuning can show useful changes with a few hundred high-quality examples, while continued pre-training needs much larger text collections. Distillation can work from unlabeled inputs. We assess your data in Discovery.
Which base models do you use?
We work with open-weight models whose licenses permit your intended commercial use. We recommend options based on your task, hardware, and license requirements, and explain the trade-offs.
What about the licenses of base models?
Each open-weight model has its own license. We review it with you, explain any restrictions on commercial use or attribution, and only use models that fit your plans.
What is your refund policy?
The Discovery fee is refundable within 7 days if work has not started. Project payments are milestone-based. See our Refund Policy for full details.
Will you sign an NDA?
Yes. We routinely sign mutual NDAs before discussing confidential details, using your template or ours.
Do you work with individual clients?
Yes. Our Private Clients service offers personal fine-tunes, local setups, and consulting hours priced for individuals and small teams.
What support do you offer after delivery?
Every project includes 30 days of email questions. Optional monthly support plans add monitoring, help with issues, and scheduled retraining.
Can the model run fully offline?
Yes. Models we deliver can run on your own hardware with no internet connection, provided the hardware is sufficient for the model size.
Can you promise a specific accuracy?
No one can honestly promise accuracy before seeing the data. Instead, we agree on measurable acceptance criteria after Discovery and report progress against them at every milestone.
Do you work outside New Jersey?
Yes. We work with clients across the United States, mostly remotely.
How do I get started?
Fill in the contact form or call us. We will schedule a free scoping call to understand your goals and suggest a next step.
Still have questions? Read about our process, see pricing, or contact us.
Not sure where to start?
Book a free scoping call. We will look at your use case, your data, and your constraints, and tell you honestly whether a custom model makes sense.
Book a free scoping call