• Created in-house application used in performance evaluation of AI models for edge device (Ambarella SoCs)

• Responsible for MobileNet-SSD based gear guard application(R1T/R1S) running on (Gen 1 (CV22) and Gen 2 (CV5)) edge device.

− Collaborated in developing gstreamer influenced publish-subscribe framework for computer vision pipeline for multi-sensor processing on edge device (Ambarella SoCs)

− Performed memory and performance optimizations for the applications running on edge platform. 
− Collaborated with CV/ML algorithm development team for model training and data collection

• Have worked extensively with buildroot and Makefiles.

• Responsible for maintaining RPC based IPC framework used for receiving raw data from IDSP (core with ThreadX) and feed it to linux application using neural net model (core with embedded linux) on hardware accelerator platform (Ambarella SoC)

• Responsible for developing image processing features using eigen, opencv (buildroot support) and ARM-NEON intrinsic on edge platform (Ambarella SoC)

• Responsible for improving the edge performance of computer vision (Mask-RCNN) and custom denoising model on ARM cores and with vendor provided tools on VPU for edge platform.

• Worked with AI/ML team in data collection and other basic training stages.

• Assisted in re-architecting the code (C++11 multithreading and CMake) using opensource message broker (Redis) and OpenCV for Cargo Recognition and Organization System application.

• Responsible for developing production code base for linux based systems using multi-threading, IPC, memory management and power management.

• Responsible for developing image processing applications using eigen, opencv and tensorflow C++ libraries.

• Resolved bugs using valgrind, gdb, cppcheck and optimized the code for improved performance.

• Developed bash scripts for unit testing and data analysis.

• Developed systemd services for deployments

• Reproduced and fixed errors associated with video decoder related codebase for MediaBox.

• Added feature to handle v4l2 events associated with resolution change for better streaming experience.

• Added features and support for video encoder.

• Implemented and maintained boot-loader.

• Implemented bare metal code for external clock detection.

• Maintained and worked on I2C, SPI and UART drivers.

• Maintained track of firmware development, bugs with JIRA.

• Maintained python test scripts for performing unit testing of developed modules.

• Involved in board bring up.

• Worked on headphone jack and accessory detection modules for the product.

• Working on many other Audio ASIC development and testing related tasks