Built for the field

AI without the cloud
Built for edge hardware

Decision Labs Edge enables on-device, data-driven decision making: forecasting and perception on the hardware you already carry and leave behind, from camera traps in remote reserves to buoys and stations with patchy uplinks. Act on what matters without waiting for the cloud.

Browse models See use cases

Loading on-device model…

Runs in your browser also on
  • iOS
  • Android
  • Embedded
  • C++
Use Cases

Built for missions that can't wait for the cloud

Turn telemetry and sensor streams into on-device decisions where connectivity is unreliable, latency is critical, and every watt counts.

Predictive Maintenance

Detect drift, flag anomalies, and act on equipment telemetry before failure: motor health, battery life, and industrial sensor arrays running fully on-device.

Toto-2 22M Chronos-2 TimesFM 2.5

Environmental Sensing

Multivariate decisions on air-quality, water, and energy sensor networks. Filter noise and trigger alerts locally before any uplink to the cloud.

Chronos-2 Toto-2 22M TimesFM 2.5

Coming soon

Urban & infrastructure Soon

Detect buildings, vehicles, solar panels, and storage tanks from satellite or aerial imagery. On-device geospatial perception weights are on the roadmap.

Building Detection Solar Panels Car Detection

Land & maritime Soon

Segment wetlands and land cover and detect ships in ports.

Wetland Segmentation Ship Detection
Model Store

On-device weights, ready to deploy

Time-series LiteRT weights for purchase and download. Geospatial perception models are coming soon.

Available now

Time-series forecasting: purchase and deploy today.

Toto-2 4M Live Time series

Smallest Toto-2 edge checkpoint. Free with registration. On-device quantile decisions at context 512.

edge 3.0 MB ctx 512

decision-labs/toto2-litert

Toto-2 22M Live Time series

Edge weight checkpoint for Datadog Toto-2.0-22m. On-device quantile forecasting at context 512.

edge 22.7 MB ctx 512

decision-labs/toto2-litert

Toto-2 313M Live Time series

Edge weight checkpoint for Datadog Toto-2.0-313M. On-device quantile decisions at context 512.

edge 317.3 MB ctx 512

decision-labs/toto2-litert

Toto-2 1B Coming soon Time series

Large Toto-2 edge checkpoint for high-capacity field telemetry. On-device quantile decisions at context 512.

edge 1.1 GB ctx 512

decision-labs/toto2-litert

Toto-2 2.5B Live Time series

Edge weight checkpoint for Datadog Toto-2.0-2.5B. On-device quantile decisions at context 512.

edge 2.5 GB ctx 512

decision-labs/toto2-litert

Chronos-2 120M Live Time series

Edge weight conversion of amazon/chronos-2. Univariate, multivariate, and covariate forecasting at context 2048.

edge 121.9 MB ctx 2048

decision-labs/chronos2-litert

TimesFM 2.5 200M Live Time series

Edge weight conversion of google/timesfm-2.5-200m. Patched transformer with host-side RevIN for long horizons.

edge 234.2 MB ctx 2048

decision-labs/timesfm2_5-litert

Coming soon

Geospatial perception weights for detection and segmentation.

Object Detection Coming soon Detection

Detect vehicles, buildings, boats, and utility infrastructure in drone or satellite imagery (WALDO30 classes).

ONNX104 MBbrowser
Zero-Shot Detection Coming soon Detection

Custom object detection from text labels when your target isn’t in a specialised detector’s class list.

ONNX204 MBbrowser
Oriented Detection Coming soon Detection

Rotated bounding boxes for ships, aircraft, and vehicles that aren’t axis-aligned in overhead imagery.

ONNX73 MBbrowser
Car Detection Coming soon Detection

Cars and small vehicles in urban, suburban, or rural imagery for traffic and parking analysis on-device.

ONNX q845.5 MBbrowser
Ship Detection Coming soon Detection

Maritime vessel detection in ports, coastal zones, and open water from satellite tiles.

ONNX q845.5 MBbrowser
Building Detection Coming soon Detection

Built structures in aerial and satellite imagery for urban development and disaster response mapping.

ONNX q845.5 MBbrowser
Solar Panel Detection Coming soon Detection

Rooftop and utility-scale solar installations in overhead imagery for energy infrastructure surveys.

ONNX q845.5 MBbrowser
Oil Storage Tank Detection Coming soon Detection

Industrial storage tanks in refineries and port facilities from satellite imagery.

ONNX q89.2 MBbrowser
Mask Generation Coming soon Segmentation

Segment contiguous regions (roads, lakes, fields, solar arrays) as precise masks on map polygons.

ONNX40 MBbrowser
Land Cover Classification Coming soon Segmentation

Categorise vegetation, urban areas, water, and other land-use types across a region of interest.

ONNX11.9 MBbrowser
Wetland Segmentation Coming soon Segmentation

Identify marsh and wetland zones for environmental monitoring and conservation planning.

ONNX q845.5 MBbrowser
Building Footprint Segmentation Coming soon Segmentation

Precise building outlines for urban mapping, planning, and post-disaster assessment.

ONNX15.7 MBbrowser
Why Edge

Decisions where the data lives.

Field hardware is fragmented: hundreds of SoC variants, NPUs, GPUs, and MCUs, each with its own compilers and runtimes. Cloud round-trips add latency and fail when uplinks drop. On-device inference is the only way to act in real time on camera traps, buoys, drones, and sensor nodes.

Google’s LiteRT is the universal on-device framework for this era: one .tflite checkpoint, accelerated across CPU, GPU, and NPU on Android, iOS, Linux, Windows, Web, and embedded targets without rebuilding your pipeline per platform.

Decision Labs Edge sits on that stack. We convert and ship production LiteRT weights for time-series and perception models so your team deploys the same artifact across the fleet, not a maze of one-off exports.

  • Unified runtime One model file, hardware-aware acceleration with CPU/GPU/NPU fallback.
  • Cross-platform by design Mobile, desktop, edge Linux, WebAssembly, and embedded from a single build.
  • Built for harsh environments Low latency and offline-first. No cloud dependency for every inference.
LiteRT .tflite
  • iOS
  • Android
  • Web
  • Linux edge
  • Embedded
  • Drones
  • IoT sensors
  • C++
SDK

One API. Every platform.

The Decision Labs Edge SDK abstracts hardware differences so your team ships faster from mobile apps to bare-metal embedded systems running in the field.

iOS

Swift bindings with Core ML and ANE acceleration. Camera pipeline integration for real-time vision on iPhone and iPad field apps.

Swift · Core ML · Vision

Android

Kotlin/Java SDK with NNAPI and GPU delegates. Background inference for rugged tablets and custom Android-based edge devices.

Kotlin · TFLite · NNAPI

C++

High-performance native runtime for Linux edge servers, NVIDIA Jetson, and custom SBCs. Zero-copy camera buffers and batch inference.

C++17 · ONNX Runtime · TensorRT

Embedded

Microcontroller targets with CMSIS-NN and custom kernels. Run detection and classification on ARM Cortex-M and RISC-V at milliwatt budgets.

C · CMSIS-NN · FreeRTOS

Linux Edge

Containerized deployment for gateway devices. gRPC and MQTT interfaces for sensor fusion hubs and drone companion computers.

Docker · gRPC · MQTT

WebAssembly

Browser and Node.js inference for dashboards, annotation tools, and rapid prototyping before hardware deployment.

WASM · ONNX.js · TypeScript

Need help shipping to the edge?

Book a one-hour support session with the Decision Labs team: LiteRT integration, weight deployment, and field rollout troubleshooting.

Browse models