DataCore is a robotics‑native backend that captures, synchronizes, and stores multi‑modal sensor streams and logs from edge fleets via a resumable QUIC‑based EdgeRelay. It offers time‑aligned slice retrieval, versioned datasets with full lineage, and REST/gRPC APIs plus Python/Rust SDKs for integration with ROS, analytics, and training pipelines, while providing RBAC security and deployment options across cloud, VPC, hybrid, or on‑prem environments.
Funding
Funding not disclosed
Founders
Product
Problem
Robotic fleets generate high‑volume, multi‑modal sensor streams and logs, but teams often lack a reliable way to capture, synchronize, and version this data across unreliable networks. Without reproducible datasets, debugging incidents and training autonomous models becomes slow and error‑prone.
Solution
DataCore is a robotics‑native backend that ingests edge‑captured streams via a resumable QUIC‑based EdgeRelay, buffers and uploads data with back‑pressure handling, and stores them in a type‑aware catalog. The platform provides deterministic, time‑aligned slice retrieval, enabling engineers to extract synchronized sensor, state, and log segments for incident replay or dataset export. It automatically packages replay bundles and creates versioned datasets with full lineage and provenance, supporting reproducible training pipelines. Integration is offered through REST/gRPC APIs and Python/Rust SDKs, allowing seamless embedding into existing ROS or custom stacks. Security is enforced via RBAC, API keys, and audit trails, while deployment can run in managed cloud, customer VPC, hybrid, or on‑prem environments to meet data‑residency requirements.
Target Audience
The primary customers are robotics teams that operate field‑deployed fleets—such as autonomous vehicle developers, warehouse automation providers, and drone operators—who need reliable data capture, incident debugging, and reproducible training datasets.
Features
- Edge capture module with timestamped multi‑modal streams, buffer‑and‑upload, and retry/resume logic for intermittent connectivity
- EdgeRelay ingestion layer using QUIC multiplexing for durable acknowledgments and priority QoS handling
- TypeAtlas semantic layer that maps payloads to stable TypeRefs, preserving schema portability across ROS and proprietary formats
- Scalable storage and indexing service that maintains raw byte stores and metadata indexes for fast slice queries
- Retrieval pipelines that generate time‑aligned slices and stream them to downstream analytics or local replay environments
- DatasetOps engine that creates versioned datasets with full lineage, provenance UI, and export connectors to training stacks
- Incident workflow producing replay bundles, annotations, and incident objects linked to original recordings
- RBAC‑based access control, API‑key authentication, and immutable audit logs for compliance
- Console UI and SDKs (Python, Rust) for programmatic access, dataset export, and operator tooling
- Flexible deployment options: fully managed cloud, customer‑controlled VPC, hybrid, or on‑prem installations with explicit retention policies