Himanshu Upreti - Curriculum Vitae

I am a Senior ML Engineer with a focus on Natural Language Processing (NLP), Computer Vision (CV), and Machine Learning (ML). I possess a Master’s degree in Computer Science and Engineering from IIT Bombay and a Bachelor of Technology (B.Tech) degree from GBPUAT Pantnagar. I have a passion for developing cutting-edge technology solutions and am proficient in Python, C++, PyTorch, Git, and have a strong track record of leading successful cross-functional teams.

Work Experience

Senior Engineer at Qualcomm CR&D

Bangalore, India | Jan 2022 - Present

• Optimizing source code of SOTA neural network for optimized peformance on AI100
• Designing and developing Graph Neural Networks projects
• Leading cross-functional teams to deliver successful projects.
• Designing optimized solutions to showcase Qualcomm’s AI100 performance in MLPerf.
• Collaborating with researchers and engineers to drive innovation and improvements in Qualcomm’s AI100 SDK.
• Mentored junior team members, fostering a culture of continuous learning and professional growth.
• Submitted two patents.

Projects

Core Team Member, Cloud AI 100 Apps SDK

• Contributed as a key member of the core development team for Qualcomm’s Cloud AI 100 Apps SDK.
• Demonstrated expertise in advancing the SDK’s capabilities, with a focus on cutting-edge Graph Neural Network (GNN) based tools.

Lead Engineer for GNN Tool:

• Led a team of three senior engineers in developing a high-impact GNN-based tool.
• Played an instrumental role in enhancing the overall functionality and effectiveness of the SDK.

Technical Contributions and Engagements:

• Delivered diverse technical solutions and presentations to the Customer Engineering team.
• Shared insights, knowledge, and innovation to drive effective adoption and utilization of the SDK’s features.

Engineer at Qualcomm CR&D

Bangalore, India | Jul 2020 - Dec 2022

• Submitted two invention diclosure forms
• Contributed to the development of computer vision applications, leading to enhanced product offerings.
• Designed and implemented a data pipeline for large-scale machine learning tasks, resulting in increased efficiency and reduced processing time.
• Developed and optimized advanced algorithms in C++ to optimize neural networks.
• Enhanced neural network operator supports in AI 100 SDK

Projects

Qualcomm Cloud AI 100 Deep Learning Inference Workflow

• RCNN Exporter

  • Led the creation of the RCNN Exporter tool, solving challenges with dynamic tensor operators in models like Faster RCNN and Mask RCNN.
  • Designed custom ONNX model generation, reimagining core components and integrating Caffe2 ops for optimal AI 100 compiler compatibility.

• ONNX Runtime execution provider

  • Developed a specialized Execution Provider for ONNX Runtime tailored for AI 100 architecture within an accelerated timeframe.
  • Enabled seamless integration of ONNX models with AI 100 hardware through the creation of the QAic Execution Provider.
  • Utilized AIC compiler and runtime libraries to develop and deliver Execution Provider in AI100 Apps SDK catering to the specific requirements of premium customers.

Qualcomm Cloud AI 100 Deep Learning Inference Tools

• Model Preparator

  • Integral part of the core team that developed the Model Preparator tool.
  • Contributed to creating an inference-friendly model generation solution, optimizing pretrained models by addressing control flow and dynamic tensor challenges.
  • Played a key role in supporting ONNX and TensorFlow models, configuring parameters via YAML format, and executing the tool effectively.

• Qaic Smart NMS

  • Implemented Smart NMS technique to optimize inference times in object detection models.
  • Orchestrated partitioning of models, capitalizing on AI accelerators for feature extraction and distributing box processing and NMS modules on the host.
  • Achieved enhanced inference speed by leveraging parallelism across AI100 and host, resulting in improved overall performance of object detection pipelines.

Research Assistant at CSE Department IIT Bombay

Mumbai, India | May 2017 - May 2020

• Held the position of Research Assistant within the prestigious CSE Department at IIT Bombay.
• Played a pivotal role in the departmental web team, assuming responsibilities for administering web servers, managing web applications, and maintaining the Computer Science and Engineering (CSE) department’s official website.
• Led interview processes for prospective web team members and departmental Research Assistants, contributing to the selection of skilled individuals to support the department’s technological initiatives.
• Orchestrated the RA admission process for consecutive years (2018, 2019), demonstrating strong organizational and coordination skills to ensure a smooth and efficient recruitment cycle.

Projects

Online Grades Evaluation System

• Led a dynamic team of three students to successfully design, develop, and implement an innovative online grade evaluation system for the Department of Computer Science at IIT Bombay.
• Utilized cutting-edge technologies including Django, Nginx, HTML, CSS, and Python to create a robust and user-friendly platform.
• Orchestrated end-to-end project lifecycle, from requirements gathering to deployment, ensuring the system met the unique needs of students, professors, and office staff.
• Designed a collaborative interface that streamlined evaluations for research-oriented courses, encompassing seminars, R&D projects, and B.Tech/M.Tech/Ph.D. thesis assessments.
• Notably, the project’s success led to its adoption by other departments as well, exemplifying its widespread value and utility across diverse academic domains.

CSE Departmental Website

• Engineered a modern backend system for the departmental website, adopting a REST API architecture.
• Successfully migrated data from an older PHP-based website to a Django-rest and MySQL-based REST architecture.
• Demonstrated technical expertise in backend development, ensuring seamless data transition while enhancing the website’s efficiency and performance.
• Enabled improved user experience and maintained data integrity through skillful data migration and the implementation of contemporary technologies.

Thesis and Research

MTech Thesis | Fall 2019- Spring 2020

Deep Neural Nets for Abnormality Detection in Medical Images | Guide Prof. Suyash Awate

• Worked on neural-net methods and algorithms for abnormality detection in various datasets of Medical Images
• Performed binary classification and analyzed one-class classification on CheXRay14 dataset
• Achieved AUC-score of 0.68 in detection of Infiltration Abnormality in Radiographs
• Further objective is to perform detection, localization and segmentation of the abnormalities

MTech Seminar | Spring 2019

Deep Learning Methods for Object Detection on Medical Images | Guide Prof. Suyash Awate

• Surveyed literature regarding various color normalization techniques in Histopathology Images
• Study and implementation of various Deep Learning architectures like ResNet and UNET for object detection
• Proposed and designed a method Patch-CNN for object detection of normal cells and infected cells in thin blood smears images
• Localized infected cells using Patch-CNN and UNET; achieved AUC-score of 0.98 and 0.96 respectively

Education

• Master of Technology (M.Tech) in Computer Science and Engineering

  • Indian Institute of Technology Bombay (Mumbai, Maharashtra, India)

• Bachelors of Technology (B.Tech) in Computer Science and Engineering

  • GBPant University of Agriculture and Technology (Pantnagar, Uttarakhand, India)

Skills

• Natural Language Processing (NLP)
• Computer Vision (CV)
• Machine Learning (ML)
• Python
• C++
• PyTorch
• Git
• Django
• Nginx
• Gradio

Hobbies

• Photography: Check out my photography blogs here.
• Fitness: Actively go to the gym to maintain a healthy lifestyle.

Contact Information

• Email: er.hupreti[@]gmail.com
• LinkedIn: Himanshu Upreti Linkedin
• GitHub: github.com/erh94
• Location: Bangalore, India