Open to research scientist roles from early 2027

Tejaswini Medi

Ph.D. candidate in machine learning at the University of Mannheim. I build generative models that get both the big picture and the fine detail right, for images and 3D shapes, and I study how to keep deep networks robust and fair under attack.

See publications Download CV Email Me
Tejaswini Medi
8publications
ECCVlatest, 2026
~10theses supervised
8venues reviewed for

About

I'm a Ph.D. candidate in the Machine Learning Group at the University of Mannheim, advised by Prof. Dr.-Ing. Margret Keuper. My research covers high-fidelity generative models in 2D and 3D, and the robustness and fairness of deep neural networks under adversarial attacks.

In 2024 I interned at Autodesk Research, where I worked on wavelet-based generative modelling for 3D shapes. Before that, I built transformer-based models for generating nested 3D shapes at Endress+Hauser.

I came to machine learning from engineering: a Bachelor's in Mechanical Engineering in Hyderabad, then a Master's in Mechatronics at the University of Siegen, where I first worked on reinforcement learning.

  • Image generation paper accepted at ECCV 2026.
  • RL-FAT accepted at BMVC 2026.
  • 3D-WAG, from my Autodesk internship, accepted at BMVC 2025.
  • FAIR-TAT accepted at WACV 2025.
  • Oral at the ICCV 2023 workshop on OOD generalization, Paris.
  • Invited talk at SFI Visual Intelligence, Norway.

Research

Three threads, one question: does the model get the details right, for every input and every group?

Frequency-aware image generation

Separating low and high frequencies so generated images keep structure and fine texture.

ECCV 2026

Coarse-to-fine 3D generation

Wavelet-guided autoregressive models and transformers that build 3D shapes, including shapes nested inside others.

BMVC 2025 · GCPR 2023

Fair and robust classifiers

Adversarial training that doesn't leave some classes far more vulnerable than others.

WACV 2025 · BMVC 2026 · ICCV-W 2023

Publications

    Full list on Google Scholar

    Experience & education

    ● research   ● education

    • 2023 – now

      Doctoral researcher

      Machine Learning Group, University of Mannheim

      Generative models for images and 3D shapes; adversarial robustness and fairness. Teaching assistant and thesis supervisor.

    • Jul – Sep 2024

      Research intern

      Autodesk Research, AI Lab

      Built 3D-WAG, a hierarchical wavelet-guided autoregressive model for fast, high-fidelity 3D shape generation. Published at BMVC 2025.

    • Mar – Sep 2022

      Student researcher

      Endress+Hauser Group

      Built FullFormer, a transformer-based implicit generative model for nested 3D shapes, including non-watertight surfaces. Published at GCPR 2023.

    • Jan – Oct 2021

      Student researcher

      University of Siegen

      Reinforcement learning for facility layout planning and dynamic job-shop scheduling.

    • 2023 – 2027

      Ph.D., Machine Learning

      University of Mannheim
    • 2019 – 2022

      M.Sc., Mechatronics

      University of Siegen
    • 2014 – 2018

      B.E., Mechanical Engineering, with distinction

      CBIT, Osmania University, Hyderabad

    Talks

    Teaching & service

    • Generative Computer Vision Models; Reinforcement LearningTeaching assistant, University of Mannheim
    • About 10 Bachelor's and Master's thesesSupervised in computer vision, generative AI and ML
    • Reviewer
      CVPRICCVECCVICLRWACVBMVCTMLRTPAMI

    Let's talk

    I'm finishing my Ph.D. in early 2027 and looking for research scientist roles in generative AI and computer vision. I'm also happy to collaborate on related research.