Bring your own models, datasets, or APIs. FloTorch supports modular experiment design through reusable, configurable blocks.

Once you have identified the experiment that works as per your defined objective, push the best-performing Agents/models or endpoints to the production pipelines with a single click.

Visual dashboards for comparing metrics across experiments using Open Source libraries such as ragas, DeepEval. You can write your own custom code to evaluate models and pipelines, as per your requirements.

Every model, model-config, agents as well as specific prompts can be versioned for A/B testing , enabling full reproducibility across experiments.

For each of the stages of GenAI pipelines such as Indexing, Embedding, Retrieval as well as Inferencing, multiple hyperparameters need to be tuned to specific values in order to achieve high accuracies at lower costs and low latency. With FloTorch integrated approach, hyperparameter tuning is done with just a few clicks.