I am an undergraduate researcher in machine learning. I hold a B.Tech in Computer Engineering from Jamia Millia Islamia, New Delhi (2022–2026), and I am completing a B.S. in Data Science and Applications at the Indian Institute of Technology Madras (expected December 2026). Since January 2026 I have been a Machine Learning Intern on the “Walk for Life” project at the Institute of Liver and Biliary Sciences (ILBS), New Delhi. There I work on clinical data pipelines and on contactless vital-sign estimation from video.

My work asks a common question: when does a model look right for the wrong reasons, and how can we measure it? In language models, I study whether models reason over nested defeasible premises in an order-invariant way, and whether their correct answers rest on correct intermediate steps. In clinical ML, I study where established screening rules fail. One example is the FIB-4 index in obese patients with MASLD (metabolic dysfunction-associated steatotic liver disease), presented as an oral talk at APASL STC 2026. I also build research software, including flwr-trees, a scikit-learn–compatible library for federated tree ensembles, and I contribute to scikit-learn and the wider scientific Python ecosystem.

Research statement

I want to build machine-learning systems whose success can be trusted for the right reasons. In both language models and clinical prediction, aggregate accuracy often hides systematic failure: a model can be accurate on average while relying on input order, surface cues, or a population it was not designed for. My goal is to develop evaluation methods that expose these failures, using controlled perturbations, subgroup analysis and step-level metrics. I also want to develop modelling approaches that remain reliable when they are deployed in clinical and other high-stakes settings. I am looking for research positions where I can pursue these questions with rigorous experimental design and real-world data.

Research interests

• Reliable reasoning in language models: order sensitivity, faithfulness of intermediate reasoning, evaluation design
• Clinical machine learning: failure modes of clinical risk scores, interpretable models, population-level evaluation
• Privacy-preserving and federated learning: communication-efficient aggregation for tree ensembles
• Physiological signal processing: remote photoplethysmography (rPPG), heart-rate variability, EEG representation learning

News

• Sep 2026: Presented “FIB-4 False Negatives in MASLD: Clinical Phenotypes and an Augmented Triage Model” as an oral presentation at APASL STC 2026 (Kumamoto, Japan), and received the APASL International Travel Award.
• May 2026: Released flwr-trees 0.1.0 on PyPI.
• Apr 2026: Pull requests merged into scikit-learn (FastICA performance, preprocessing validation fix) and scikit-bio (import-time reduction). See Open Source.
• Jan 2026: Joined the Institute of Liver and Biliary Sciences, New Delhi, as a Machine Learning Intern.
• Dec 2024: Winner, Smart India Hackathon 2024 Grand Finale (problem statement 1680, NTRO).

Education

• B.S. in Data Science and Applications, Indian Institute of Technology Madras, 2023 – Dec 2026 (expected)
• B.Tech in Computer Engineering, Jamia Millia Islamia, New Delhi, 2022 – 2026. CGPA 8.70 / 10

Selected honours

• APASL International Travel Award, APASL STC 2026
• Winner, Smart India Hackathon 2024 (Grand Finale, NTRO)
• Rank 98 of 32,000+ teams (top 0.3%), Amazon ML Challenge 2024
• Winner, HackJMI 2024
• Qualified, National Standard Examination in Physics (NSEP) 2021; qualified JEE Advanced 2021

A full CV is available here (PDF).