Enlight delivers comprehensive tools designed for enterprises to create and update AI applications, saving valuable time and resources. Through its API and user-friendly interfaces, businesses can develop custom AI solutions with ease. The platform leverages advanced knowledge transfer algorithms and neural network architecture to ensure highly accurate and performant results, even for users without a machine learning background or with minimal training data. Enlight supports various custom model training methodologies, including leveraging pre-trained Clarifai Models, utilizing Enlight QuickTrain for rapid transfer learning, or employing Enlight DeepTrain for ground-up custom embedding training to maximize performance. Additionally, it allows for the import of third-party models, offering maximum flexibility. The platform also provides robust evaluation tools that automate the testing and validation process, including 5-split cross-validation, to ensure the fastest time-to-accuracy for custom models. Users can evaluate and showcase different model versions, predict performance against designs before production, and adjust thresholds for recall and precision rates to optimize accuracy. Beyond training, Enlight enables models to analyze data streams for faster real-world responses, automate business processes by scanning vast amounts of data, simplify the search and classification of unstructured data by focusing on key concepts, and analyze operational patterns to track business performance and trends.

Features & Benefits

• Clarifai Models
  • Leverage pre-trained models with millions of expertly labeled inputs, providing out-of-the-box AI readiness.
• Enlight QuickTrain
  • Rapidly build custom AI models by transferring learnings from Clarifai models or your own deep-trained models.
• Enlight DeepTrain
  • Maximize model performance by training custom embeddings from the ground up.
• 3rd Party Model Integration
  • Import and utilize your own third-party models within the Clarifai platform for maximum flexibility.
• Automated Model Evaluation
  • Take the guesswork out of performance assessment with automated testing and validation processes, including 5-split cross-validation and random subset generation.
  • Evaluate and showcase all trained model versions.
  • Predict against designs and test performance before production.
  • Adjust thresholds for recall and precision rates to optimize accuracy.
• Real-World Data Stream Analysis
  • Train models to analyze data streams for faster real-world response and improved situational awareness.
  • Identify safety issues on a factory floor.
  • Detect forest fires in remote locations.
  • Manage inventory and keep store shelves stocked.
  • Trigger events and notifications.
• Business Process Automation
  • Automate tasks by training models to scan millions of data points much faster than human capabilities.
  • Moderate social media and user-generated content (UGC).
  • Automatically label and organize training data sets.
• Simplified Unstructured Data Search
  • Train models to focus on key "concepts" within unstructured data, making it easier to search, organize, classify, and understand.
  • Process millions of images and videos for classification and labeling.
• Operational Analysis & Pattern Isolation
  • Create models that isolate patterns in business systems and processes to track and analyze performance.
  • Count visitors to a new business location based on promotions.
  • Learn how a product is trending in social media over time.