R.A.D. Lab AI positions itself as a Spatial Intelligence Platform. Its core idea is to turn “raw physical signals” into “queryable semantic knowledge.” It aims to ingest data such as 3D scans, documents, audio, and RF signals, extract structured semantic scene graphs through domain adapters, and then let users perform reasoning and queries in natural language. Its current focus is on a 3D Spatial Search SaaS, while it is also looking for research collaborators in robotics, spatial computing, signal processing, and embodied intelligence.
Based on the main content, its architecture is split into a general core and domain adapters. The general layer includes user profile storage, a Graph Store + Vector Index, and a LangGraph-based RAG Agent. The graph store contains typed nodes, edges, weights, and entity vectors, using pgvector for semantic retrieval and NetworkX for graph traversal. The adapter layer handles specific signal-processing tasks, such as Open3D for point clouds, PyMuPDF for documents, and librosa for audio. For semantic annotation, it mentions CLIP/SigLIP, CLAP, and NER.
The clearest use case is uploading a 3D spatial scan and asking questions in natural language, such as whether there is enough wheelchair clearance between a table and a door. The planned 3D adapter supports PLY, OBJ, PCD, and E57, with capabilities including point-cloud segmentation, RANSAC plane detection, zero-shot object labeling, and spatial relationship edges such as proximity, containment, and clearance. The document adapter targets entity extraction for people, organizations, topics, dates, and more. The limitation is that the product is still in Active Development: the 3D and document adapters are under development, CAD/BIM and video/real-time streams remain planned, and no performance metrics or real-world case results have been disclosed.
The page does not provide a free tier, trial, pricing, payment methods, deployment options, or data privacy details, so there is not enough information for commercial procurement. On integrations, it only discloses file/data formats and internal technical components; there is no visible public API, SDK, or enterprise system integration documentation.
The main strengths are a clear technical direction, a focus on the difficult “signal-to-semantics” problem in robotics and spatial intelligence, and the combination of graphs, vector retrieval, and RAG to solve complex relational queries. The drawbacks are that the product is still not very mature, and support, compliance, pricing, and accessibility are not transparent. It is better suited to research teams, robotics/spatial AI startups, and early partners willing to co-develop, rather than enterprises looking for a mature plug-and-play SaaS.
The main content does not state whether it is accessible from mainland China, whether a Chinese interface is available, or whether local payment methods are supported, so china_access can only be marked as unknown. For Chinese team projects, it is recommended to first confirm website connectivity, cross-border data transfer requirements, and payment methods. Possible alternatives include combining locally deployed point-cloud processing, knowledge graphs, vector databases, and RAG frameworks.
⚠ This review is compiled from public sources and does not constitute a purchase recommendation. Verify all facts on the vendor's official site. Verify on radlab.ai official site.
radlab.ai is an United States AI Apps provider. TG4G tracks its product information, an overall rating of 7.0/10, and a China-accessibility score of Workable. Click "Visit Official Site" to reach radlab.ai directly.