Subho (Subhojeet Pramanik)

Subho (Subhojeet Pramanik)

MSc Graduate

RLAI, University of Alberta

" I think; therefore I am. " — René Descartes.

I work in reinforcement learning and artificial intelligence. I completed my MSc in Computing Science at the University of Alberta and was co-supervised by Adam White and Marlos Machado; affliated with RLAI Lab and Alberta Machine Intelligence Institute (Amii).

My research interests lie broadly in reinforcement learning, representation learning and continual learning. In my MSc thesis, I proposed a recurrent alternative to the transformer’s self-attention mechanism, which offers context-independent inference cost and parallelization over an input sequence. The proposed approach called the Recurrent Linear Transformer was shown to outperform state-of-the-art transformers and recurrent neural networks in partially observable reinforcement learning problems, both in terms of computational efficiency and performance. ( Thesis URL, ICLR Submitted Paper)

I have worked in several industry positions in machine learning. During my MSc, I interned at Huawei Research Edmonton, applying reinforcement learning to neural network operator fusion. Previously, I had worked with IBM Cloud as an ML Engineer (around 2 years) and collaborated with IBM Research on various research projects in representation learning and deep learning. I also helped deploy several machine learning algorithms at scale in IBM and Kone.

Contact: spramanik [at] ualberta [dot] ca, email [at] subho [dot] in

Interests

  • Reinforcement Learning
  • Representation Learning
  • Deep Learning

Education

  • MSc in Computer Science (thesis based, Fully funded), 2021 - 2023

    University of Alberta

  • B.Tech in Computer Science and Engineering, 2015 - 2019

    Vellore Institute of Technology

Experience

 
 
 
 
 

Researcher Intern

Huawei

Nov 2022 – Jun 2023 Edmonton, Canada
 
 
 
 
 

AI/ML Developer

IBM, Cloud

Sep 2019 – Jul 2021 Bangalore, India

Primarily assigned as an AI/ML Developer for IBM App Connect:

  • Part of the newly minted AI team responsible for Research → Development → Productionization of Mapping Assist feature for IBM Appconnect, Cloud Pak for Integration release.
  • Introduced a novel unsupervised technique for database schema matching, and deployed the same using ONNX+FastAPI+Docker pipeline.

Actively collaborating with IBM Research:

  • Worked towards a AAAI, 21 submission to learn a generic document representation by leveraging multi-page, multi-modal text & visual information in real-world PDF documents. (Mentors: Dr. Sameep Mehta, Shashank Mujumdar, Hima Patel)
 
 
 
 
 

Intern

IBM, Watson IoT

Jan 2019 – Jul 2019 Bangalore, India

Intern at the IBM Watson TRIRIGA Building Insights team.

  • Researched on methods to formulate and predict HVAC efficiency and the factors that cause HVAC degradation. Proposed a decision tree based approach for HVAC failure detection and root-cause analysis. (Mentors: Alex J Joseph, Sattwati Kundu)
  • Actively collaborated with Priyanka Agrawal from IBM Research towards a research publication: OmniNet: A unified architecture for multi-modal multi-task learning
 
 
 
 
 

Founding Partner

Cognibit Solutions LLP

Jan 2018 – Jan 2019 Remote & Finland

Responsibilities include:

  • January 2018 - December 2018 (Remote contract work with Kone Corporation, Finland): Achieved record accuracy in predicting elevator door failures and scheduling service needs from elevator sensor logs.
  • January 2019 (On-site collaboration with Kone R&D team and IoT team at Finland): Deployed the developed predictive maintenance solution on IBM Kubernetes platform.
 
 
 
 
 

Visiting Researcher

Kone Corporation

May 2017 – Jun 2017 Hyvinkää, Finland

Selected amongst hundreds of competitors in Kone IBM hackathon for a two month sponsorship to Kone in Finland as a visiting researcher.

  • Performed predictive modeling of customer callouts with elevator log data using LSTM which acheived record accuracy.
  • Performed statistical modeling & diagnostics of Kone’s german elevator data extracted from the Remote Maintenance Program which yielded commercially valuable insights.

Mentors: Dr. Olli Mali, Jani Hautakorpi