ModelPlus Integrator provides comprehensive automation and low-code/no-code support throughout the AI application development workflow, particularly for the creation, training and integration of multiple AI models, deployment, and operation of analytical engines on the cloud.
Integrator ships with two user portals. First, the Workflow Studio (the front-end) provides an intuitive and declarative way to construct complex AI applications that can include multiple models and interleaved control/data flows. through automatic code generation and verification, the application design is compiled to workflow specifications. Then, the Workflow Engine (the back-end) deploys and executes the workflows on the best infrastructure proven from industry, including container, real-time data pipelines and source management system that guarantee the system’s robustness and interoperability.
Compared with the creation of applications mainly through manual integration today, ModelPlus Integrator offers the following significant advantages:
No-code orchestration tools verified across multiple scenarios, precisely constructing complex visual AI applications, eliminating development costs.
Formal method verification from academia ensures the correctness and reliability of the constructed AI systems.
Modular system design enables continuous accumulation and updates of high-quality AI models, cross-scenario reuse, significantly reducing costs.
Cloud + terminal + edge integrated hardware-neutral deployment mode, without platform lock-in, providing unlimited customization flexibility for customers.

While the ModelPlus Integrator aims to eliminate tedious coding from the application lifecycle, the ModelPlus Insight goes one step further, unlocking autonomy by building domain specific AI assistants - which take instructions on how to use the video and answer any questions from the user. Both instructions and questions are assigned to agents and ultimately transformed to workflows executed by ModelPlus Integrator.
There has never been a better time to build Gen AI powered systems, as new building blocks and paradigms emerge literally every day. At the same time we have never been equipped with so many diverse and extraordinary choice of tools: with RAG (Retrieval- augmented generation) one can create a pipeline from data retrieval, semantic search to action and micro-optimize every step, whereas other large multi-modal models allow decision making in a zero-shot fashion. There are many desirable goals that are not always compatible and lead to necessary product trade-offs. For example, large models (such as GPT, Sora, Midjourney) can generate as much sophisticated content as humans can imagine but lack control and extensibility. on the other hand, rule -based systems such as a firewall or antivirus defense system are customizable at the finest granularity but are poor at generalization and lack creativity. To address these challenges, ModelPlus Insight wants to provide both generative/creative power and extensibility, at the cost of having a less simple architecture. The following two philosophical aspects distinguish our product from others in this domain.
Explicit Alignment
We want to build products that can be used in mission critical tasks such as surveillance and business analytics. We believe our AI should be contextualized and explicitly aligned to such tasks, as opposed to using universal capability and providing best effort. The main product challenge of ModelPlus Insight is to make the alignment efficient, which is addressed by providing a powerful domain extension mechanism that defines metadata, models, prompts and test cases.
Data Modeling
We believe the keystone in alignment is data modeling, which connects the cognitive models with world models in a domain. to generate trustable actions, we need to provide a way for the users to verify and be convinced of generated insights. Toward this goal, ModelPlus Insight internally creates an (extensible) metadata model and a compact lakehouse that captures the most essential data elements. Mapping cognitive problems to the data domain also has the additional benefit of leveraging tons of mature optimization for performance and scalability.