August 1st 2025

Bojai is built for ML researchers, students, and engineers who want full control over their model pipelines without heavy DevOps. This release gives you total control over how you build, train, and launch ML pipelines.

💡 Bojai Pipelines Got a Major Upgrade

Simpler. Smarter. Totally Yours.

Bojai 2.0 lets you shape each pipeline stage to fit your workflow. Mix and match models, edit training logic, and control outputs—all from one unified interface.

👉 Want to try it? Explore the Docs

💾 Save Your Work. Pick Up Where You Left Off.

Faster. Safer. Right Where You Left Off. Training a model can be time-consuming—but now, you don’t have to start from scratch every time. Bojai lets you save your training sessions (left GIF), including their model parameters, training loss, and evaluation. Later, you can load it instantly by name (right GIF) and resume where you left off. This is especially helpful for long-running sessions, hyperparameter tuning, or collaborative workflows.

🔁 Need to pause training? Now you can.

Try Session Saving →

📊 Log and Visualize Your Training Like Never Before

Transparent. Insightful. Built-in. Bojai now includes full logging and visualization support during training and deployment. You can track loss, accuracy, and evaluation scores at every stage—helping you debug, optimize, and understand your models with confidence. Whether you're training locally or deploying remotely, Bojai gives you the tools to see what's working.

Logging is handled automatically inside trainer.py using the Logger class. Visuals are generated via the Visualizer, which reads from the log and supports multiple views. Training and validation stats are tied to epochs, while final evaluation is plotted separately for clarity.

# Logging during training
self.logger.log(epoch=epoch, loss=loss, train=score, valid=val_score)

# Logging final evaluation score
self.logger.log(eval_score=final_score)

🧠 Your model has something to say—see it clearly.

Learn more →

🧠 Use the Data Processing Agent to Adapt Any Dataset

Beta. Flexible. LLM-powered.