Your AI. Your data. Your control.
The platform that lets companies and individuals build their own foundational AI models — far more efficient than conventional training techniques.
Train on your own hardware or on Sapiens infrastructure. No dependence on third-party models, no giving up your data.
CPU
is all it takes to train your model
no GPU cluster required
The proprietary Sapiens architecture evaluates multiple internal outputs before answering, instead of returning the first generation from the core algorithm. That delivers higher quality with far less training data and far less compute.
In practice: training that is viable at CPU level. Foundational models stop being the privilege of whoever owns GPU clusters and come within reach of your company.
Four architectural decisions that change who can train a foundational model — and under what conditions.
Developed from scratch in Brazil. Not a wrapper over third-party transformers: a training and internal-evaluation approach of our own.
Train on Sapiens infrastructure when you want speed, or inside your own network when the data cannot leave the building.
Compute cost reduced to the point of not needing dedicated clusters. The new era of AI training starts on machines you already own.
The training data is yours. The resulting model is yours. No handing proprietary knowledge over to someone else's model.
Pick the approach that fits your use case and your volume of data.
Start from a trained model
Fine-tune an already-trained model with your own data. Today the base can be any model you trained yourself on Builder — fine-tuning on top of the proprietary SAPI models is coming soon. Your model inherits the Schizophrenic AI architecture, including the internal multi-model evaluation.
Build your Foundation Model
Use the proprietary SAPI architecture as the foundation and train an entirely new model on your data. No dependence on third-party models — 100% yours, with Schizophrenic AI efficiency from the first iteration.
Resume training an existing model
Continue training a model you already have, folding in new data without starting from scratch. Today the base needs to be a model you trained yourself on Builder — continued training on top of the proprietary SAPI models is coming soon.
curl -sN -X POST "http://localhost:8080/fine-tuning" \
-H "Content-Type: application/json" \
-d '{
"base64_dictionary": {},
"dataset_path": "./datasets/dataset.json",
"model_name": "base_model.model",
"output_name": "fine_tuned_model",
"progress": true
}'With Sapiens Builder, your company trains personalized AI models with its own data — simple, fast and efficient.
Connect and organize your company's data: files, systems and internal databases.
Use the platform to prepare, train and tune your model with total control.
A personalized model, trained on your data and optimized for your business.
Integrate the model into your apps, systems and workflows.
Reinforce learning with new data; keep the model always current and relevant.
More precision, more efficiency, more autonomy. Real results for your business.
Sensitive data stays within your perimeter throughout training. Nothing has to cross your organization's boundary.
No vendor lock-in. Your model is not switched off, deprecated or repriced by someone else's decision.
The efficiency of the architecture puts training your own model within reach of teams that would never have had the budget for it.
The trained model is an asset of your company — versionable, auditable and improvable at your own pace.
Upload your data, pick a training path and have your first model running on the Sapiens architecture.
Create now