cv

Table of contents

General Information

Full Name Akshit Achara
Email acharaakshit@gmail.com
Current Role PhD Student, King's College London
Research Areas Computer vision, representation learning, shortcut learning, interpretability

Experience

  • Oct 2024 – present

    London, UK
    PhD Student
    King's College London
    • Investigate how learned models form, align, and misuse predictive representations, with emphasis on failure modes hidden under aggregate metrics for medical image segmentation.
    • Develop testable approaches for localising shortcut reliance in biased vision models.
    • Study how demographic attributes become encoded in medical imaging models and how this affects robustness, fairness, and generalisation.
  • Oct 2020 – Jul 2024

    Bangalore, India
    Research Engineer, Applied AI Research
    GE Research / GE Healthcare
    • Developed vision models for structural-defect detection, including the prediction of near-invisible cracks that were difficult to identify manually.
    • Built patient-level representations for longitudinal oncology records to support summarisation, family-history extraction, entity linking, structured information extraction, and clinical-trial matching.
    • Led development of an end-to-end clinical-trial matching system spanning trial ingestion, eligibility structuring, patient entity linking, and ranking.
    • Represented GE Healthcare at G20 events in India, presenting applied AI initiatives to cross-stakeholder audiences.

Education

  • Oct 2024 – present

    London, UK
    PhD, Biomedical Engineering and Imaging Sciences
    King's College London
  • Aug 2016 – Jul 2020

    Pilani, India
    Integrated M.Sc. Mathematics
    Birla Institute of Technology and Science (BITS Pilani)

Selected Publications

  • Achara, A., Manickam, V., Day, T., Puyol Anton, E., Hammers, A., and King, A. P. Discovery and Spatial Characterisation of Multiple Shortcut Groups for Auditing Vision Model Bias. arXiv, 2026.
  • Gaintseva, T., Achara, A., Slabaugh, G., Deng, J., and Elezi, I. EquiSteer—Cross-Attention Steering Towards a Fairer Text-Guided Image Generation. ECCV, 2026.
  • Achara, A., Gaintseva, T., Mahaut, M., Chakraborty, P., Johansson, V. S., Barsbey, M., Rodolà, E., and Crisostomi, D. Multi-Way Representation Alignment. ICML 2026; ReAlign at ICLR 2026.
  • Avci, M. Y.*, Achara, A.*, King, A., and Cardoso, J., for the Alzheimer's Disease Neuroimaging Initiative. Understanding Sources of Demographic Predictability in Brain MRI via Disentangling Anatomy and Contrast. FAIMI at MICCAI, 2026.
  • Achara, A., Yathathugoda, Y., Byrne, N., Antonelli, M., Puyol Anton, E., Hammers, A., and King, A. P. Right Regions, Wrong Labels—Semantic Label Flips in Segmentation under Correlation Shift. CAO at ICLR, 2026.
  • Achara, A., Triantafillou, P., Puyol-Antón, E., Hammers, A., and King, A. P. Localising Shortcut Learning in Pixel Space via Ordinal Scoring Correlations for Attribution Representations (OSCAR). arXiv, 2025.
  • Achara, A., Puyol Anton, E., Hammers, A., and King, A. P. Invisible Attributes, Visible Biases—Exploring Demographic Shortcuts in MRI-based Alzheimer's Disease Classification. FAIMI at MICCAI, 2025.
  • Achara, A. and Chhabra, A. Watching the AI Watchdogs—A Fairness and Robustness Analysis of AI Safety Moderation Classifiers. NAACL, 2025.
  • Achara, A., Sasidharan, S., and N, G. Efficient Biomedical Entity Linking—Clinical Text Standardisation with Low-Resource Techniques. BioNLP at ACL, 2024.

Patents

  • Achara, A., Goravar, S., Sasidharan, S., and Kanamarlapudi, A. Clinical context-centric natural language processing solutions. U.S. Patent No. 12,632,671, issued May 19, 2026.
  • Sasidharan, S., Achara, A., Kanamarlapudi, A., Goravar, S., and Bhoi, K. Systems and methods for intelligent medical editors. U.S. Patent Application 18/906,528, 2025.
  • Ghose, S., Sasidharan, S., Cho, S., Achara, A., Mullick, R., Kanamarlapudi, A., Ginty, F., Bhardwaj, A. R., Mani, S., and Sarachan, B. Natural language based clinical recommendation system. U.S. Patent Application 18/468,593, 2025.

Awards and Service

  • UKRI-funded DRIVE-Health CDT Scholarship, King's College London, 2024–2028.
  • Best Team Award, LOGML 2025.
  • Top 10% Reviewer, NeurIPS 2025.
  • Reviewer: NeurIPS (2024, 2025, 2026), ICML (2025), AISTATS (2026).

Skills

Programming Python, C, C++, R, Java
Frameworks PyTorch, Lightning, Transformers
Research Computer vision, representation learning, shortcut learning, interpretability