About Me

Hello 👋, I’m Mayank Kumar Pal. I’m from New Delhi, India 🇮🇳. Working as the Senior Engineer at Qualcomm, Bangalore in the AI SW Team for Automotive Buisness Unit.

My interest spans area such as On-Device Machine Learning, Large Language Models(LLMs), Reinforcement learning, and Embedded Systems. I received my Bachelor’s and Master’s degree from IIIT-Delhi, Indraprastha Institute of Information Technology (IIIT-Delhi) in Electronics and Communications Engineering (ECE), where I was advised by Prof. Sanjit K. Kaul and Prof. Saket Anand.


Work Experience

  • Aug 2020 - Present

    [Full-Time] Working with Qualcomm Neural Processing (QNN) SDK team, where we optimize trained models for on-device inference and performance on Qualcomm devices. Currently, I focus on the automotive business unit working on next-gen automotive chipset from Qualcomm. Worked on identifying the issues in model performance and suggest various optimization to improve the KPIs

  • Dec 2019 - May 2020

    [Internship] Developed a framework in C/C++ to accelerate linear algebra and signal processing algorithms using OpenCL for heterogeneous devices. The framework was designed to utilize the DSP and GPU to improve the performance of the algorithms. Performance of the system was evaluated using Andreno GPU and Intel FPGAs.

  • May 2019 - Jul 2019

    [Internship] During my internship, made significant contributions to the cloud infrastructure. Secondly, automated the build, test, and deployment workflows. By creating robust pipelines for Continuous Integration and Continuous Deployment, we streamlined the development process. Additionally, containerized our applications using Docker images, eliminating cross-platform compatibility issues.

  • May 2018 - Jul 2018

    [Internship] Worked in Philips’ innovation team in Bangalore, where I developed a pipeline to systematically interact with authoring kit APIs. This pipeline matched user query intents using Keras-trained Intent Classification models. Additionally, I extracted named entities (such as names, locations, and organizations) from unstructured text to anonymize personal data using Keras models.