Deep Learning Studio

The Deep Learning suite is six pages: Deep Learning (browse all 16 models) → Studio (orchestrate data) → Architecture Builder (design) → Training Monitor (train) → Experiments (compare) → Deploy (ship). It is fully no-code, yet you can export runnable PyTorch / TensorFlow / ONNX code at any step.

From the Studio you can

• Install a one-click sample dataset (14 ready-made sets auto-create a project and provision data)
• Upload your own — CSV, Excel, JSON, Parquet, image folders/ZIPs, audio folders, or graph CSVs
• Or label images in-browser with the built-in annotator on the Training page — draw boxes, regions, or labels and train, no pre-labelled data needed
• Pick a Quick-Start template, or ask the AI architecture advisor for a dataset-specific recommendation
• Browse the Model Marketplace and click Adopt to open the Architecture Builder seeded with the closest matching template

14 model families you can train

• Core: Tabular (MLP / LSTM / GRU / Transformer), Vision (CNN / ResNet / EfficientNet), Segmentation (U-Net), Multimodal (image + tabular)
• Specialized: NLP text classification, Audio classification, Causal inference (TARNet), Graph (GCN), Federated (FedAvg), Meta-learning (Reptile), Generative (DCGAN), Diffusion (DDPM), Object detection (Faster R-CNN), Foundation (ViT + LoRA)