The collaboration between Intel and Intel Ignitel alumni Deci has significantly improved AI performance through the integration of Deci’s Deep Learning Development Platform with Intel technology.
Deci’s Deep Learning Development Platform gives developers and data scientists a new way to use deep learning models to their fullest potential by making it easier to build, optimize, and deploy models that are faster and more accurate.
Intel’s association with Deci started in 2019, focusing on the optimization of deep learning inference on Intel Architecture (IA) CPUs. The relationship was bolstered by Intel Ignite, which introduced Deci to customers and facilitated meetings with high-level executives. Intel Ignite played a crucial role in shaping and driving the partnership between the two entities. Given the challenges small companies face in partnering with larger corporations, Intel Ignite is assisting Deci in streamlining their collaboration process and helping Intel understand optimal strategies for working with smaller startups. Moreover, Intel Ignite continues to provide significant value to Deci on a weekly basis.
In 2020, Deci made headlines by submitting impressive inference results to the MLPerf v0.7 benchmark suite, achieved with its proprietary AutoNAC (Automated Neural Architecture Construction) engine and 2nd Gen Intel® Xeon® Scalable processors. By leveraging the Intel® Distribution of the OpenVINO™ toolkit, Deci managed to cut the inference latency of the ResNet-50 deep neural network model by a staggering 11.8x and boost its throughput by 11x, all while maintaining model accuracy within 1%.