To showcase a developer engagement featuring its facial processing and facial recognition technologies, the Qualcomm Developer Network asked venTAJA to write a case study on Magisto’s CamCrew app. They also commissioned a demonstration guide for the app, from which a related video script evolved.

Magisto CamCrew Video Script

Qualcomm engineers heard from developers keen to accelerate matrix multiply (MM) operations for convolutional neural networks in their deep learning (DL) apps. The Qualcomm Developer Network engaged venTAJA to work on an application note as a series of blog posts about accelerating deep learning on Qualcomm’s Adreno GPU.

Matrix Multiply for Deep Learning on Adreno GPUs –

Part 1: OpenCL Optimization • Part 2: Host Code and Kernel

As processor performance increases, developers can move more of the computational work of artificial intelligence (AI) from the cloud down to mobile devices. venTAJA worked on learning resources and blog posts for the Qualcomm Developer Network to introduce programmers to running workloads more efficiently among the CPU, GPU and DSP on mobile devices at the network edge.

Neural Processing learning resources • Ollama simplifies AI inference with open-source models

Most developers find out quickly that machine learning algorithms and frameworks perform slowly on the CPU, so they turn to the digital signal processor (DSP) for better performance. The Qualcomm Developer Network hired venTAJA to work on a series of blog posts describing how the Hexagon SDK eases the transition to programming machine learning applications on the DSP.

Use C++ in Your Machine Language Programming with Hexagon SDK 3.1