R. Suresh

AI Research Engineer | VFS Global

profile.jpeg

New Delhi, India

I build production AI systems at VFS Global processing over 1 million documents daily across 110+ countries. My research focuses on uncertainty quantification and reasoning reliability, specifically when and why language and vision-language models fail to know what they don’t know.

B.Sc. in Programming & Data Science, IIT Madras. Always happy to talk research or engineering, so feel free to reach out.

news

Apr 07, 2026 Posted SELFDOUBT to arXiv, uncertainty quantification for reasoning LLMs via the Hedge-to-Verify Ratio that outperforms Semantic Entropy at about 10x lower cost. Link
Apr 05, 2026 Posted Don’t Blink to arXiv, evidence collapse during multimodal reasoning. Link
Jan 11, 2026 Published an article on prompt ordering in vision-language models: Why Your VLM Prompts Are Backwards (And How to Fix It)
May 02, 2024 Joined AI @ VFS as lead AI engineer

latest posts

Jan 11, 2026 Why Your VLM Prompts Are Backwards (And How to Fix It)
Mar 16, 2024 No-Nonsense Guide to LangChain
Nov 17, 2023 From Images to Insights: 3-Tiered Data Extraction from images with OCR and Large-Language Models

selected publications

  1. db.png
  2. sd.png
    SELFDOUBT: Uncertainty Quantification for Reasoning LLMs via the Hedge-to-Verify Ratio
    Satwik Pandey, Suresh Raghu, and Shashwat Pandey
    2026
    Shows better uncertainty quantification performance than Semantic Entropy at about 10x lower cost
    arXiv