Ved Umrajkar

Ved Umrajkar

Joint Secretary, DSG IIT Roorkee

Education

  • BS-MS, Mathematics and Computing, IIT Roorkee (2021–2026)

    Research Interests

  • Machine Learning
  • Computer Vision
  • Applied Statistics
  • Time Series Analysis
  • Financial Mathematics

Publications

  • Ved Umrajkar, Aakash Kumar Singh. (2025). Detection Limits and Statistical Separability of Tree Ring Watermarks in Rectified Flow-based Text-to-Image Generation Models. ICLR 2025.

    Projects

  • Voice Cloning
    Developed automated voice clones using Tortoise-TTS with a Streamlit application for real-time custom voice generation.
  • ForecastPro
    A time series forecasting package implementing SOTA deep learning models (LSTNet) and statistical models (ARIMA, SARIMAX) with evaluations on UCI and Yahoo Finance datasets.

  • GANs vs Mixture Models
    Comparative analysis of different approaches to image generation using DCGAN, WGAN, and GMMs with Expectation Maximization, finding GMM-generated images were more diverse but lacked GAN sharpness.

  • Portfolio Optimization
    Implemented multi-risk measure optimization on S&P-500 stocks using Sharpe ratio and Monte Carlo simulation for forecasting portfolio returns.

Awards

  • Expert on Kaggle (Top 1%)
  • Two-time Amazon ML Challenge Finalist with 17th Rank in 2023 (10,000+ candidates)
  • Silver Medal at the Inter-IIT Tech Meet 13.0

  • All India Rank 342 in KVPY Fellowship Examination

    Follow

    Kaggle | CodeForces | GitHub —