Present, Senior Software Engineer @ AMD, Toronto

At the ROCm™ profiling team at AMD, Toronto, we are working towards advancing the suite of tools used to profile, benchmark and optimize AI workloads on AMD Instinct™ Data Center accelerators.

Over the last year I have been contributing to the rocprofiler-compute component of the open source ROCm™ stack. For more details, see this post This tool enables performance profiling of GPU kernel code used in various HPC and AI workloads to help optimize the kernel code by providing key insights on its performance characteristics on AMD’s Data Center GPUs such as MI 100, 200, 300 and 350 series of GPUs.

It collects raw performance counters from various IP blocks of the GPU and derives performance metrics such as Speed-Of-Light metrics, wavefront occupancy, roofline (compute vs memory bound), memory hierarchy bandwidth and data transfer, GPU pipeline stalls, scheduling efficiency, instruction mix and other useful metrics. Moreover, it allows ROCtx based tracing to correlate source code with associated performance metrics. Other features include Program Counter (PC) sampling to investigate instructions stalls, roofline empirical benchmark and filtering options for fast and efficient profiling. To collect large number of raw performance counters, the tool leverages application replay and optionally provides iteration multiplexing ability to prevent application replay for large workloads with counter accuracy trade-off.

Senior AI Researcher @ AMD, Toronto

As part of the User Experience Group at AMD, Toronto, I was working on the cross-roads of Computer Vision, Natural Language Processing and Machine Learning to improve gaming experience using Radeon Software. In particular, I worked on projects such as Game Event Detection, Visual Quality Assessment, Game Graphics Recommendation Engine and Efficient Retrieval Augmented Generation.

Masters in Applied Computing @ University of Toronto

I completed the Masters in Applied Computing program at the University of Toronto with specialization in Artificial Intelligence in December 2023. These four courses shaped my masters experience:

  • Internship at AMD, Markham on Computer Vision for Video Games
  • Deep Learning and Neural networks
  • Statistical Machine Learning
  • Computational Imaging
  • Data Science Methods

Diploma in Data Science @ IIT Madras

In order to focus in the Artificial Intelligence domain after working in the software engineering sector for four years, I decided to pursue a diploma. I completed an online diploma program in Data Science from Indian Institute of Technology, Madras on July 2022. Notable courses in this program were:

  • Statistics
  • Machine Learning Foundations and Techniques
  • Linear Algebra, Probability Theory and Combinatorics

Senior Member Technical @ D. E. Shaw India

I worked with the Systems team from 2018 to 2022. The Systems team developed and managed internal applications to provide software infrastructure to the entire firm. As a full stack DevOps Engineer, I had to develop, test and maintain internal applications. I also worked as on-call escalation support for maintaining our Windows and Linux servers and applications running on them. Some notable examples of internal applications include:

  • Hardware Inventory management
  • Permission management
  • Credential manager which runs locally for fast access to resources
  • Slack app integration for critical server alerts

Bachelors in Computer Science @ IIT Patna

I completed Bachelors degree in Computer Science from Indian Institute of Technology, Patna on May 2018. Following courses influenced me the most as part of this journey:

  • Bachelors Thesis in Natural Language Processing using Deep Learning
  • Artificial Intelligence
  • Data Structures and Algorithms
  • Computer Networks
  • Operating Systems

During my second year, I completed a 3 months long summer internship at Cigniti, Hyderabad, India. Here, I explored cloud services to develop machine learning pipelines using drag and drop UI and deploy them in cloud. This is when I completed the famous Coursera machine learning course by Andrew Ng.

My third year saw me work as a summer intern at Elsevier, Chennai, India. I was able to secure this internship due to recommendation by my bachelors thesis supervisor, Prof. Asif Ekbal. The work I had done at AI-NLP-ML lab in IIT Patna under his supervision was useful in this internship. I created a pipeline to extract the contents from a corpus of research papers in medical domain. Clean and extract TF-IDF features from the contents. Built an unsupervised clustering model using Non-Negative Matrix Factorization to extract meaningful cluster of topics from the corpus.