Ph.D. Candidate · Statistics · University of Michigan

Soham Bakshi

Research

Modern data science and machine learning present new challenges that calls for rigorous theoretical frameworks to analyze complex, high-dimensional data and to ensure reliable inference from data driven methods. My research aims to devise innovative statistical methodologies that provide robust inferential guarantees, with applications in causal inference and contemporary data-science. My current research interests can be organized around the following themes:

  1. Explore randomization as a tool to facilitate flexible and tractable inference for a wide-range of data-driven methods

  2. Develop novel statistical frameworks to advance theoretical insight into the complexities arising in today’s AI & ML systems

Publications & Preprints

  1. Bakshi, S., Gao, L., Gao, Z. and Panigrahi, S. (2026). Flexible Inference for Winners with Conditional Validity. Preprint, arXiv:2607.18545. Under review, Journal of the Royal Statistical Society: Series B.

  2. Bakshi, S. and Chakraborty, S. (2026). From Collapse to Improvement: Statistical Perspectives on the Evolutionary Dynamics of Iterative Training on Contaminated Sources. Preprint, arXiv:2602.10531. Under review, NeurIPS 2026.

  3. Bakshi, S. and Panigrahi, S. (2026). Classification Trees with Valid Inference via the Exponential Mechanism. Preprint, arXiv:2511.15068. Under review, Bernoulli.

  4. Bakshi, S., Huang, Y., Panigrahi, S. and Dempsey, W. (2024). Inference with Randomized Regression Trees. Preprint, arXiv:2412.20535. Under revision, Biometrics.

  5. Bakshi, S., Dempsey, W. and Panigrahi, S. (2024). Selective Inference for Time-Varying Effect Moderation. Preprint, arXiv:2411.15908. Second-round revision, Journal of the American Statistical Association.

  6. Bakshi, S. and Maity, S. (2023). Bayes Classifier Cannot Be Learned from Noisy Responses with Unknown Noise Rates. In International Conference on Learning Representations (ICLR), Tiny Papers Track. OpenReview.

  7. Bhattacharyya, T., Bakshi, S. and Das, A. (2022). How to Approximate Irrational Numbers Nicely? Preprint, arXiv:2206.12579.

  8. Bakshi, S. (2022). On the Set of Points of Continuity and Maximally Discontinuous Extensions. Resonance 27 131–142. doi:10.1007/s12045-022-1298-1.

Work in Progress

  1. Conditional Inference after Gibbs-Randomized Selection: Exact Conditional Laws and Selection-Conditional Asymptotics beyond Finite Candidate Sets. Manuscript in preparation.

  2. Inference-Aware Randomization for Conditional Selective Inference: Convex Channel Design under Selection Regret. Manuscript in preparation.

Software

  1. SI-CART — selective inference with trees.

  2. LLMCollapse — iterative training on contaminated sources.

Selected Talks

  1. Randomized Trees with Conditional Inference. Joint Statistical Meetings (JSM), Boston, MA, USA, 2026. Slides.

  2. Selective Inference for Time-Varying Moderated Effects. Statistical Analysis of Multi-Outcome Data (SAM), Salzburg, Austria, 2024. Slides.

  3. Non-Identifiability of Bayes Classifier. Joint Statistical Meetings (JSM), Toronto, Canada, 2023. Slides.

  4. Geometry in Statistics. Math Club, Indian Statistical Institute, Bangaluru Centre, India, 2022. Slides.

  5. Hopf-Rinow’s theorem and Geodesic Completeness. Visiting Students Research Program (VSRP), School of Mathematics, TIFR, India, 2021. Slides.

Awards & Fellowships

  1. Rackham Conference Travel Grant, University of Michigan, 2026.

  2. Rackham International Travel Grant, University of Michigan, 2023–2024.

  3. Visiting Students Research Program, selected with honorarium, School of Mathematics, Tata Institute of Fundamental Research, Mumbai, India, 2021.

  4. Summer Research Fellowship Programme (offered), Indian Academy of Sciences, Indian National Science Academy and National Academy of Sciences, India, 2021.

  5. Regional Mathematical Olympiad qualifier and participant, Indian National Mathematical Olympiad Training Camp, Indian Statistical Institute, Kolkata, India, 2018–2019.

Service

  1. Reviewer, International Conference on Machine Learning (ICML), 2026.

  2. Reviewer, Journal of the Royal Statistical Society: Series B, 2026.

  3. Core Organizing Committee member, Michigan Student Symposium for Interdisciplinary Statistical Sciences (MSSISS), University of Michigan.

  4. Member, Institute of Mathematical Statistics (IMS) and American Statistical Association (ASA).