Solution
Transform your data to insights, using tailored cloud solutions built on a generic AI agent platform.
Hybrid approach
Traditional machine learning and specialised AI agents combine in concert to sequentially analyse data, create reports and present calls to action, storing periodic assessments for further assessment
Agentic technology
Tool-enabled networks of specialised agents solve complex problems, analyse images, diagnose sensor data, assess reports, using a rich toolkit and a flora of LLMs, assembling to become virtual experts
Configured to purpose
Agents built and hosted on Aidalos can be instructed at different levels, familiarised with your assets, sensor configuration and requirements, and be given access to your own proprietary tools
Orchestrated workflow
A collection of numerical methods and generalist and specialist AI agent, are orchestrated in workflows, providing you with combined data analysts and domain experts, on need or 24/7
Building virtual specialists on our generic core
Aidalos Core
Specialized and customized agents built on our agentic framework can be instructed at different levels, familiarized with your assets, sensor configuration and requirements
Aidalos Applications
Aidalos can be configured-to-purpose manifestations of Aidalos Core, enabling you to configure your own agent and connect your data streams, resulting in context-aware, domain-specific, sets of agents.
Benchmarking
Background
Generative AI, as inherently fuzzy and probabilistic without strict separation between true and false, is difficult to assess in terms of quality
We established five real-life technical-oriented tasks to be solved in the context of provided images, and gave them to Aidalos, ChatGPT 4o and ChatGPT 4o Deep Research
Evaluation
The task descriptions and the corresponding answers were given to a third party, a well-reputed AI-powered research assistant, and assessed according to standard (see e.g.1-4 on this list) criteria:
• Stick to the point. Are all parts of the response relevant?
• Completeness. Does the response seem to completely answer the question?
• Accuracy. Does the response accurately answer the questions?
• Hallucination. How much does the response contain constructed facts?
Tasks
A: Analyse this specific image in terms of production complexity. |
B: Check if reinforcement placement complies with general requirements and evaluate the castability. Can you suggest improvements? |
C: Analyse the photo and give an assessment. What should be done? |
D: Where should we put sensors on to monitor this bridge? |
E: The following is a plot from an anomaly detection campaign. Do a deep analysis and summarize the situation. What could explain the observations? The task description and answers are found here. T |
Results
Aidalos performed well, granted an "A" score (you can find the evaluation results here):
|