cv

Research Interests

  • I am a PhD candidate in AI at Queen Mary University of London. My research focuses on understanding intermediate representations in foundation models and developing methods to analyse and steer model behaviour. I am particularly interested in representation learning, interpretability, and control of generative systems. I have a strong background in machine learning and mathematics, along with prior research experience in industry.

Education

Experience

  • 2024 — 2025
    Research Intern , Huawei Noah's Ark, London, UK

    Researched intermediate representations in vision-language generative models and developed methods for steering generation and image-processing behaviour. First-authored three papers during the internship, including one accepted to ICLR 2026.
  • 2021 — 2022
    Data & AI Scientist , Philips Innovation Labs, Moscow, Russia

    Developed novel AI-based techniques for Philips medical-device software. Founded and led a research reading group focused on recent advances in AI for medicine.
  • 2020 — 2021
    Researcher , Huawei Russian Research Institute, Moscow, Russia

    Led research on AI-based face recognition, improving overall pipeline performance by 2.5%. Coordinated a Huawei–MIPT collaboration on domain adaptation methods for face recognition, including research review and project communication.
  • 2018 — 2020
    Research Assistant , LAMBDA lab, Moscow, Russia

    Conducted research on generative models for high-energy physics. Designed and implemented an original pipeline for synthetic particle-event generation. Co-authored two scientific papers, including one published in an ICLR workshop.
  • 2017
    Research Intern , Yandex, Moscow, Russia

    Applied reinforcement learning to optimize metrics for a seq2seq vocalization task. Combined multiple seq2seq models using ideas from the Actor-Mimic algorithm and showed that RL improved model quality. This work formed part of my bachelor's thesis.

Selected Publications

  • 2026
    Georgii Aparin, Tatiana Gaintseva. A Geometric Account of Activation Steering through Angle-Norm Decomposition. arxiv preprint. [paper]
  • 2026
    Tatiana Gaintseva, Andrew Stepanov, Ziquan Liu, Martin Benning, Gregory Slabaugh, Jiankang Deng, Ismail Elezi. MidSteer: Optimal Affine Framework for Steering Generative Models. ICML 2026. [paper, code]
  • 2026
    Akshit Achara, Tatiana Gaintseva, Mateo Mahaut, Pritish Chakraborty, Viktor Stenby Johansson, Melih Barsbey, Emanuele Rodolà, Donato Crisostomi. Multi-Way Representation Alignment. ICML 2026. [paper, code]
  • 2025
    Tatiana Gaintseva, Chengcheng Ma, Ziquan Liu, Martin Benning, Gregory Slabaugh, Jiankang Deng, Ismail Elezi. CASteer: Steering diffusion models for controllable generation. ICLR 2026. [paper, code]
  • 2024
    Laida Kushnareva, Tatiana Gaintseva, German Magai, Serguei Barannikov, Dmitry Abulkhanov, Kristian Kuznetsov, Irina Piontkovskaya, Sergey Nikolenko. AI-generated text boundary detection with RoFT. COLM 2024, Outstanding Paper Award. [paper, code]

Additional Training

  • 2026
    ARENA 8.0 , London, UK

    AI safety research engineering programme focused on mechanistic interpretability, model evaluations, reinforcement learning, and alignment-relevant ML engineering.
  • 2025
    LogML Summer School , London, UK

    Contributed to the research project "Representational Alignment for Universal Spaces", which won the Best Project Award at LogML 2025. The resulting work was accepted to the ICLR Re4-Align 2026 workshop.
  • 2016 — 2018
    Master's level program , Yandex School of Data Analysis

    Moscow, Russian Federation. Completed coursework in machine learning, deep learning, applied statistics, and big data.

Selected Teaching

Skills

  • Programming
    • Python, PyTorch, NumPy, pandas, scikit-learn, Git
  • Machine Learning
    • Deep learning, transformers, large language models, diffusion models, generative models, reinforcement learning, representation learning
  • Technical Focus
    • Interpretability, analysis of intermediate representations, steering and control of foundation models, AI safety
  • Mathematical Background
    • Linear algebra, probability theory, differential geometry, statistics, optimization