Trellis
Intelligence
Platform
The Trellis Intelligence Platform (TIP) is built for flexibility and explainability.
Capable of ingesting data from any source, working with multiple content types, and pushing results to practically any software application or enterprise system.
Internally, TIP is used by Trellis Data to train and refine AI models, and is also made available as a standalone product to our customers, providing a seamless option for integrating AI capabilities into existing software and enterprise platforms, particularly for government agencies and large-scale operations.
We want to help organisations leverage secure and explainable AI solutions so they can innovate, create efficiencies, and master their decisions.
TIP is designed to support a wide range of AI model types.
Generative AI for knowledge mastery across vast data sets, generating text and code in a secure environment.
Speech Models for transcription, translation, sentiment/stress analysis, and keyword alerts.
Vision models for object detection, person of interest, scene analysis, and scene text reading
Modal Chaining technology is a key advantage of TIP and is ideally suited for creating AI solutions to complex workflows.
Benefits of Chaining AI Models
Improve Accuracy: Reduce the risk of errors by using multiple models to confirm and validate results.
Gain Deeper Insights: Combine different models to gain a better understanding of the data and its relationships, revealing new patterns and connections.
Increase Confidence: Use multiple models to cross-check and verify results, increasing confidence in the accuracy of the findings.
Make Systems More Reliable: Use multiple models to handle potential failures and adapt to changing conditions, making systems more robust and resilient.
Adapt and Refine: Use prior results to adjust focus and provide a more flexible and layered approach, allowing for continuous improvement and refinement.
Some Real-world Opportunities
Facial Recognition in Crowds: A facial recognition model identifies an individual in a crowd. Chaining this model with a gait analysis model could add an extra layer of confidence to the identification. By verifying the person's walking pattern against known characteristics, we can reduce the risk of false positives and increase accuracy.
Surveillance of Sensitive Areas: An Open Vocabulary Detection (OVD) model detects individuals in restricted areas of a facility. Chaining this with a Visual Language Model trained on specific behavioral analysis model could then identify suspicious patterns, and alert security to a potential breach or other possible issue.
Inventory Tracking within a Warehouse: A computer vision model tracks inventory items in a warehouse, detecting locations and quantities. Chaining this with a natural language processing model that reads labels and descriptions, and a machine learning model that optimizes storage and retrieval, enables data-driven inventory management and improved operational efficiency.
Machine learning that can evolve with your enterprise
The Trellis Intelligence Platform features continuous improvement processes that enable our customers to refine and update their models over time, ensuring they remain accurate and effective.
Model Training: TIP was originally conceived to train machine learning models from scratch and is ideally suited for retraining and fine-tuning applications.
Model Evaluation: TIP also provides tools for model evaluation, enabling customers to assess the performance of their models and identify areas for improvement.