Sayak Mukherjee

Doctoral Researcher

Hey there, Thanks for stoppping by!

I am Sayak Mukherjee (সায়ক মুখার্জী), a doctoral researcher at the Cognitive Affective Technology (CAT) Lab at TU Delft working under the guidance of Bernd Dudzik and Tom Viering. My research focuses on developing data-efficient, context-sensitive, and privacy-preserving multimodal AI systems for automatic affect prediction—paving the way for Artificial Empathic Agents. Current deep learning approaches struggle with data efficiency and in capturing situational, cultural, and personal contextual subtleties, limiting their real-world effectiveness. My work aims to bridges insights from cognitive psychology and (probabilistic) machine learning to model human cognitive-affective processes more effectively, thereby addressing these key challenges.

Beyond affective computing, I am passionate about generative modelling and test-time adaptation of pre-trained models. Holding a Master’s in Computer Science, with one year of research experience at the Computer Vision Lab, TU Delft and three years of experience as an application developer at Oracle, I bring a unique blend of industrial expertise and deep research curiosity to the future of human-centered AI.

📚 Research Interests

  • Probabilistic ML
  • Generative Modelling
  • Computer Vision
  • Affective Computing
  • Self-supervised Learning
  • Multimodal AI

🎓 Education

  • MSc Computer Science
    Delft University of Technology

  • BTech Information Technology
    West Bengal University of Technology

 

🧑‍💻 Experience

Doctoral Researcher | Delft University of Technology | Mar 2025 - Present

  • Focus: Data-Efficient ML for Context-Sensitive Affective Computing.

Researcher | Delft University of Technology | Sep 2023 - Feb 2025

  • Worked on generative models with a focus on diffusion models.
  • Explored training-free alignment of diffusion models.
  • Proposed an alignment approach based on importance sampling.
  • Collaborated with researchers from MIT, Shell AI Research and Google.

Application Developer | Oracle | Aug 2018 - Jul 2021

  • Collaborated with financial stakeholders to gather requirements.
  • Customized Oracle FLEXCUBE platform for regional banking needs.
  • Developed 40+ functionalities tailored to Latin American banking sector.
  • Utilized Java, Spring, Hibernate, SOAP API, and JavaScript.
  • Deployed solutions on Oracle Database and Oracle Weblogic Server.

Check out my CV →

📝 Selected Papers

  • “CoDe: Blockwise Control for Denoising Diffusion Models.”
    Anuj Singh, Sayak Mukherjee, Ahmad Beirami, Hadi Jamali-Rad (2025).
    In ICLR Workshop on Deep Generative Models: Theory, Principle, and Efficacy

  • “MAPL: Model Agnostic Peer-to-peer Learning.”
    Sayak Mukherjee, Andrea Simonetto, Hadi Jamali-Rad (2024).
    In ArXiV

See all publications →