Aadya Arora

Aadya Arora

Data Scientist at Microsoft. My research sits at the intersection of vision-language models, medical imaging, and generative AI, building efficient and interpretable systems for real-world healthcare.

Incoming PhD · Johns Hopkins University (EECS) · Medical AI

About

If you've ever used PowerPoint Copilot to clean up a messy slide, there’s a small chance you've run into some of my work. I was involved as a Data Scientist on that team at Microsoft. Before that, I completed my undergrad at IIT Gandhinagar.

During my time there, I got to work on some exciting research. My work focused on vision-language models, diffusion-based generative AI, and few-shot medical image classification. Through research internships at IISc Bangalore, the University of Bath, and IIIT Hyderabad, I gained experience with diffusion transformers, attention-based image editing, and anomaly detection in autonomous driving.

But right now, I’m taking a big step that I have been incredibly eager to start: my PhD at Johns Hopkins University, where I am fortunate to be advised by Prof. Vishal Patel. I'm genuinely excited to jump into research and see where this next chapter takes me.

Thanks for stopping by! Feel free to look around or reach out if you'd like to say hi.

Publications

MedFocusCLIP: Improving Few-Shot Classification in Medical Datasets using Pixel-Wise Attention
IEEE ICASSP 2025

A. Arora, V. Namboodiri

Integrated SAM-based pixel attention with CLIP for interpretable and robust few-shot medical image classification.

Flagship IEEE SPS Conference · CORE A · h5-index: 110

WavShadow: Wavelet-Based Shadow Segmentation and Removal
ICVGIP 2024

S. Jain*, A. Arora*, V. Vekaria, K. Gandhi, S. Raman (*Equal Contribution)

Wavelet-enhanced pipeline for shadow detection and removal, combining frequency-domain analysis with deep learning.

Premier Indian Conference in Computer Vision · h5-index: 18

TEMPEST: A Machine Learning Framework for Battery Thermal Prediction and Chemistry Selection
IEEE TIA 2026

A. Arora*, S. Patil*, P. Bhardwaj (*Equal Contribution)

LSTM-based framework for predicting internal battery temperature and recommending optimal chemistry for EV and grid storage.

Q1 Journal · Impact Factor: 4.4 · h5-index: 78

Battery ETHOS: Electro-Thermal Model for State Estimation
Under Review · IEEE COMPEL 2026

S. Patil*, A. Arora*, P. Bhardwaj (*Equal Contribution)

Electro-thermal modelling approach for accurate battery state estimation, extending the TEMPEST framework.

Q2 Journal · Impact Factor: 1.5 · h5-index: 25

Experience

Microsoft IDC May – July 2024

Data Scientist Intern · Received Full-Time PPO

  • Worked with the Copilot team for PowerPoint; integrated AI to improve structured presentation creation.
  • Built datasets by extracting and structuring text & formatting info from slides in JSON format.
  • Developed data pipelines to generate diverse formatting examples; improved model performance through error analysis, leading to a PPO.
IISc Bangalore · VAL Lab Dec 2024 – Apr 2025

Research Intern — Diffusion Transformers

  • Layer-wise ablation of Diffusion Image Transformers (DiTs); studied how features evolve for structure vs. style.
  • Designed unified attention masks to reduce prompt leakage in multi-region text-to-image generation.
  • Ran LoRA and DreamBooth experiments to understand editing behaviour at different transformer depths.
University of Bath May – July 2024

Research Intern — Vision-Language Models

  • Developed AdaptHIPIE: lightweight adapters into HIPIE for open-vocabulary referring segmentation.
  • Achieved 85.15 mIoU on RefCOCO, improving visual-text alignment with CLIP features.
  • Concurrently developed MedFocusCLIP, rethinking annotation-efficient methods for the medical domain.
IIIT Hyderabad · CVIT Lab Dec 2023 – Jan 2024

Research Intern — Autonomous Driving

  • Analysed corner cases in Indian traffic datasets (IDD, CODA) for unstructured environments.
  • Used class-agnostic RPNs to capture anomalous instances beyond predefined categories.
  • Built a pipeline using semantic segmentation & common-class suppression to surface safety-critical anomalies.

Achievements

PhD Offer
Johns Hopkins University

Fully funded PhD position in EECS to specialise in Medical AI.

Industry
Microsoft PPO — Data Scientist

Full-time Pre-Placement Offer at Microsoft IDC with the highest compensation in a tech role from IITGN.

Publication
ICASSP 2025 · Cash Prize

Only B.Tech student in the batch to publish at IEEE's flagship SPS conference. Awarded ₹25k by IITGN.

Service
IEEE Reviewer

Official reviewer for IEEE Signal Processing Letters (2025–present); 4 technical peer-reviews to date.

Invitation
Microsoft Research India Summit 2025

One of few students nationwide selected for India's premier annual AI academic summit.

Competition
Inter-IIT Tech Meet — Top 3

Ranked Top 3 for Research Paper Poster across all 23 IITs in EE and CSE domains.

Teaching
Teaching Assistant × 3 Courses

TA for Machine Learning, Probability & Data Visualization, and Principles of EE.

Outreach
Featured in Quill Magazine

Featured in the April 2026 edition; Problem Statement Setter for IITGN's flagship HackRush '26.

Skills

Languages

Python C++ C MATLAB

Deep Learning

CNNs ViTs CLIP Diffusion SAM LSTMs

Frameworks

PyTorch TensorFlow OpenCV mmDetection Azure OpenAI

Domains

Computer Vision Medical Imaging Generative AI Document AI