DECI AI
Ramat Gan, Israel
$100M
VALUATION
About
Israeli deep learning platform for model optimization. Automatically optimizes neural networks for faster inference and deployment efficiency.
About Deci AI
Israeli deep learning platform for model optimization. Automatically optimizes neural networks for faster inference and deployment efficiency.
AI Focus Areas
- Neural architecture search
- Model compression and quantization
- Edge and server inference optimization
- Hardware-aware deep learning
- Automated ML performance tuning
Key Products
- Deci AI model optimization platform
- Runtime container and Edge SDK for optimized inference
Market Position
Before its acquisition, Deci competed in the model optimization space by offering a platform that used AI to optimize AI, automatically generating architectures that delivered better accuracy–latency tradeoffs on specific hardware targets. Its pitch to enterprises was the ability to run more accurate models with lower compute costs, particularly on edge and constrained devices. By abstracting away low-level hardware tuning, Deci reduced time-to-production for deep learning teams and competed with both open-source tools and commercial offerings like Neural Magic or built-in vendor compilers. NVIDIA’s acquisition underscores the strategic importance of such tools in improving GPU utilization and broadening the range of viable AI workloads.
AGI Relevance
Scaling toward AGI is constrained not only by model quality but also by efficiency and energy use. Deci’s focus on automatically crafting efficient architectures is directly relevant to AGI’s compute budget: systems that can find better models under fixed hardware constraints enable more experimentation and deployment. Its approach—meta-learning over model search spaces—also echoes broader AGI research into systems that can improve themselves or optimize other networks. The acquisition by a leading GPU vendor suggests that future AGI platforms will likely embed such optimization layers deeply within their stacks, making them an integral part of how models are trained and served globally.
Investment Highlights
Deci raised $9.1M in seed funding in 2020, followed by a $21M Series A and a $25M Series B led by Insight Partners in 2022, bringing total disclosed funding above $55M before being acquired by NVIDIA in 2024. Deal terms were not publicly disclosed in major news outlets.
Tags
- Model Optimization
- Inference
- Edge AI
Recent News
No recent news for Deci AI