SmooSense provides a plug‑and‑play visual interface that turns folders and files—whether stored locally or on S3—into intuitive workspaces for browsing and exploratory analysis of multimodal data. It supports common modalities such as images, audio (with auto‑generated mel‑spectrograms), video, 3D objects, and PDFs, and lets users add custom visualizations via iframes, consolidating folder browsing, tables, and galleries in one tool.
Funding
Funding not disclosed
Founders
Product
Problem
Analyzing large collections of multimodal files—such as images, audio, video, 3D models, PDFs, and JSON—often requires moving data into specialized tools, writing custom scripts, or using separate applications for each format, which hampers rapid exploratory analysis.
Solution
SmooSense provides a plug‑and‑play visual interface that operates directly on local or S3 folders, eliminating data ingestion steps. Its unified workspace combines a folder browser, table view, and gallery, allowing users to browse, filter, and drill down into multimodal datasets with a single click. Built‑in support automatically renders common modalities (e.g., image masks, audio spectrograms, video playback, 3D object viewers) and highlights missing values. Queries are translated into optimized DuckDB SQL that runs in parallel on the files themselves, delivering instant slice‑and‑dice performance without a separate query engine. Users can extend the platform by embedding custom visualizations via iframes, enabling deeper analysis tailored to unique data types.
Target Audience
Primary users are data scientists, machine‑learning engineers, and researchers who work with heterogeneous media datasets and need an interactive, code‑friendly environment for exploratory analysis.
Features
- Direct operation on local or S3 directories with no data ingestion or external database required
- Integrated visual workspaces that merge folder browsing, tabular data, and media galleries
- Automatic rendering for images (with masks/annotations), audio (mel‑spectrograms), video, 3D objects, PDFs, JSON, and other formats
- Real‑time filtering and distribution charts that update instantly via parallel DuckDB SQL execution
- One‑click drill‑through to view associated samples and detailed metadata
- Missing‑data visualization that shows percentage gaps alongside actual data distributions
- Extensible architecture allowing custom visualizers to be embedded as iframes
- Open‑source SDK for developers to integrate the IDE into existing code workflows