Anurag Dutt

Anurag Dutt

Ph.D. Candidate

My name is Anurag Dutt, a Ph.D. student at Stony Brook University, currently working at the PACE lab under the guidance of Dr. Anshul Gandhi at PACE Lab. My academic pursuits revolve around the intriguing intersection of Distributed Systems and Deep Learning, with a focus on model serving, placement, scheduling and parallelism across GPUs. Beyond academia, I find solace in pursuits like badminton, anime, high-fantasy novels and playing tabla, an Indian percussion instrument.

DOWNLOAD CV

Recent and Upcoming Events

June, 2024

  • Interning at Observe.inc as a research intern in the AI team for the summer of 2024.

May, 2024

  • Very excited to present our paper/poster "GAMMA: Graph Neural Network-Based Multi-Bottleneck Localization for Microservices Applications" at WWW 2024 on 15th May, 2024 in Singapore. See you there!!!!!

Feb, 2024

  • Our paper titled "GAMMA: Graph Neural Network-Based Multi-Bottleneck Localization for Microservices Applications" has been accepted at WWW 2024 in top 5% of papers.

Dec, 2023

  • Successfully defended my Ph.D. Research Preliminary Examination on "Evaluating the energy impact of device and workload parameters for DNN inference on edge".

Oct, 2023

  • Presenting our work on "Evaluating the energy impact of device and workload parameters for DNN inference on edge" at IGSCC 2023 in Toronto, CA. See you there!!!

Aug, 2023

  • Our paper titled "Evaluating the energy impact of device parameters for DNN inference on edge" has been accepted at IGSCC 2023 (co-located with MICRO).

July, 2023

  • Attending USENIX ATC 2023 and OSDI 2023 in Boston, MA. It was a great learning experience!!!!!

Aug, 2022

  • Starting my PhD. !!!!!

Dec, 2021

  • Our paper titled "B-MEG: Bottlenecked-Microservices Extraction Using Graph Neural Networks" has been accepted at ICPE 2022.

Experience

Sr. Associate (Information Technology & Data Science)

Roc Capital LLC

Led machine learning infrastructure efforts at a hedge fund, optimizing deployment and serving of models for pricing real estate derivatives. Leveraged MLFlow, Docker, Kubernetes, and Spark to streamline model development and deployment pipelines, ensuring scalability and reliability. Designed model performance monitoring and maintenance systems using Prometheus, Tensorboard, and Grafana.

Jan, 2019 - Aug, 2019, May, 2021 - Aug, 2022

Data Science Intern

Thomson Reuters

Spearheaded the creation and rollout of deep learning pricing models tailored specifically for exotic credit default swaps. Established robust infrastructure pipelines using Jenkins for efficient deployment and serving. Implemented real-time monitoring solutions leveraging Prometheus and Fiddler, enabling continuous tracking of model performance metrics. Additionally, integrated TensorRT for accelerated model inference, optimizing computational efficiency and reducing latency in real-time pricing calculations.

May 2018 - August 2018

Research Associate

Reserve Bank of India

Designed and implemented a scalable data warehousing and visualization platform using Hive, Spark, Tableau, and MongoDB, handling portfolio data for 14.6 million investors over 10 years with 236 billion records. Orchestrated efficient data processing and transformation, ensuring high performance and reliability. Integrated MongoDB for flexible storage and retrieval of unstructured data components. Utilized Tableau for intuitive visualizations, empowering stakeholders with actionable insights for informed decision-making.

June 2014 - May, 2017

Education

Stony Brook University

PhD. in Computer Science - ML Systems

GPA: 4.0

August 2022 - Present

Stony Brook University

Masters. in Applied Mathematics & Statistics (Minor in Computer Science)

GPA: 3.85

August 2019 - May 2021

Columbia University

Masters. in Operations Research & Financial Engineering

GPA: 3.3

August 2017 - December 2018

Birla Institute of Technology, MESRA

Bachelors. in Engineering (Civil)

GPA: 3.67 (WES)

Publications

Conference Publications

2024

  • Social Learning via Bayesian Inverse Reinforcement Learning: Learning from and about a Learner

    2024 ACM Web Conference (WWW)

    Alexandra Ortmann, Anurag Dutt, Christian Luhmann

  • GAMMA: Graph Neural Network-Based Multi-Bottleneck Localization for Microservices Applications

    2024 ACM Web Conference (WWW)

    Gagan Somashekar, Anurag Dutt, Mainak Adak, Tania Lorido Botran, Anshul Gandhi

2023

  • Evaluating the energy impact of device and workload parameters for DNN inference on edge

    Stony Brook University (SBU)

    Anurag Dutt

  • Evaluating the energy impact of device parameters for DNN inference on edge

    Work-in-progress track, 2023 14th International Green and Sustainable Computing Conference (IGSCC)

    Anurag Dutt*, Sri Pramodh Rachuri*, Ashley Lobo, Nazeer Shaik, Anshul Gandhi, Zhenhua Liu
    *joint first-authorship

2022

  • B-MEG: Bottlenecked-Microservices Extraction Using Graph Neural Networks

    2022 The 13th International Conference on Performance Engineering (ICPE)

    Gagan Somashekar, Anurag Dutt, Rohith Vaddavalli, Sai Bhargav Varanasi, Anshul Gandhi

2020

  • SMOOTH-GAN: Towards Sharp and Smooth Synthetic EHR Data Generation

    2020 The 18th International Conference on Artificial Intelligence in Medicine (AIME)

    Sina Rashidian, Fusheng Wang, Richard Moffitt, Victor Garcia, Anurag Dutt, Wei Chang, Vishwam Pandya, Janos Hajagos, Mary Saltz & Joel Saltz

Awards & Certifications

  • Chairman’s fellow in Department of Computer Science, SBU for 2022, 2023 and 2024
  • Student travel grant to attend USENIX ATC ’23
  • Student travel grant to attend IGSCC ’23
  • Data Analytics 360 certification from Cornell University - 2018
  • Awarded National Talent Scholarship Exam Scholarship from Government of India -2013