NinjaLABO provides an AI Compression as-a-Service platform (tinyMLaaS) that lets edge AI developers upload TensorFlow, PyTorch, or ONNX models and receive automatically pruned, quantized, and weight‑clustered versions optimized for microcontrollers, smartphones and other low‑power devices. The service delivers up to 90% size reduction while preserving accuracy, and offers a web dashboard, API and CLI tools for seamless integration into development pipelines, reducing latency, cloud costs, and hardware requirements.
Funding
Funding not disclosed
Founders
Product
Problem
Deploying AI models on edge devices often requires large model sizes and high computational resources, leading to increased cloud costs, latency, and limited feasibility for low-power hardware.
Solution
NinjaLABO offers an AI Compression as-a-Service platform (tinyMLaaS) that lets users upload their existing AI models and receive optimized, size‑reduced versions suitable for on‑device execution. The service applies automated model pruning, quantization, and architecture tuning to maintain accuracy while dramatically lowering memory and compute demands. Compressed models can be downloaded and deployed directly to microcontrollers, smartphones, or other constrained hardware, reducing inference latency and eliminating the need for continuous cloud inference. NinjaLABO provides a web interface and API for batch processing, performance reporting, and integration into existing development pipelines.
Target Audience
Primary customers are edge AI developers, IoT device manufacturers, and mobile app teams that need to run inference locally on resource‑constrained hardware.
Features
- One‑click model upload supporting common formats (TensorFlow, PyTorch, ONNX)
- Automated pruning, quantization, and weight clustering to achieve up to 90% size reduction
- Compatibility checks and code generation for popular microcontroller SDKs (e.g., TensorFlow Lite for Microcontrollers, ARM CMSIS‑NN)
- Performance dashboard showing accuracy retention, latency, and memory footprint before and after compression
- RESTful API and CLI tools for integration into CI/CD workflows
- Secure, encrypted storage of uploaded models and generated artifacts