Ameet Rao← Back to experience

2025

AI/Robotics Research; UC Berkeley mentor

AI Researcher — Co-author

UpstreamQA: A Modular Framework for Explicit ReasoningOpen PDF ↗

Highlights

  • Co-author of AI research applying a novel vision-language model framework with applications for robotics
  • Processed NExtQA dataset videos into formats for Gemini and ChatGPT
  • Ran object detection and scene context experiments, with and without reasoning
  • Demonstrated performance gains up to +9% in LLM-Match accuracy
  • 1 of 17 papers accepted worldwide to VLM4RWD (NeurIPS 2025), presented in Mexico City in November 2025

Other experiences