Jul 2021 – Jul 2022 · 1 yr

• Cadence AI Boost NNE 110 - Neural Network Accelerator
• Played a major role in enabling multiple TensorFlow network models for audio/voice applications like keyword detection and noise suppression
• Performance analysis of LSTM operator within various networks on the accelerator
• Developed frameworks in bash to build, profile and test networks

Jul 2018 – Jun 2021 · 2 yrs 11 mos

• Dolby MS12 Multi-stream Decoder (Dolby Atmos) and Dolby Harmonious (Home Audio) on Tensilica HiFi3 DSP
• Responsible for Dolby's Audio Post-processing component and Integration Framework
• Performance optimization of low-level fixed point intrinsics
• Developed a Python framework for objective testing of MS12 for critical testcases
• Created multiple scripts to speed up code development and debug time. These are now used across the teams
• TensorFlow Lite for Microcontrollers on Tensilica HiFi5 DSP
• Kernel optimizations of CNN operator (Depthwise Separable Convolution)

Jul 2015 – Jun 2018 · 2 yrs 11 mos

• Dolby MS12 Multi-stream Decoder on Tensilica HiFi3 DSP
• Worked on API development, performance optimization, testing and integration of various components:
• Encoders for Dolby Digital Plus and MAT formats
• Decoders for HE-AAC, AC4 and MAT formats
• PCM renderer and Multi-stream Mixer

Jul 2013 – Jun 2015 · 1 yr 11 mos

• Performance and memory optimization of AAC-ELD and DRM codecs from FHG on Tensilica HiFi3 DSP
• API development for Dolby Digital Plus Decoder on Tensilica HiFi2 DSP

Jun 2012 – Dec 2012 · 6 mos · Pune Area, India

• Variable Length Encoder/Decoder
• Developed a tool to encode and decode patterns using Huffman coding. Using this tool resulted in great performance gain on Tensilica HiFi2/HiFi3 for AAC codecs

May 2011 – Jul 2011 · 2 mos · Hyderabad, Telangana, India