Arnav Choudhry

About Arnav Choudhry

I am an Applied Scientist working at the intersection of advanced machine learning, generative AI, and cyber-physical infrastructure. With a PhD focused on Machine Learning for Cyber-Physical Infrastructure, my research has historically spanned everything from simulating robotic platforms and multi-modal transit to quantifying flight energy risk for autonomous UAVs.Today, as part of Amazon’s Last Mile Science team, I bridge the gap between cutting-edge methodology and massive operational scale. I architect geospatial models and robust safety metrics that serve millions of daily routing requests. By integrating Vision-Language Models (VLMs) and GenAI \"world models\" into spatial modeling, I aim to improve GPS accuracy, simulate complex routing outcomes, and proactively manage real-world driver risk. My core philosophy is to be a rigorous scientist first, focusing on interpretability, causal investigation, and trustworthy AI, while remaining relentlessly production-aware.Beyond the algorithms, my life is grounded in community and the outdoors. I love to travel, sketch, and play squash, and I care deeply about the natural world, regularly volunteering for trail maintenance, tree planting, and park cleanups. Recognizing the profound impact that access to opportunity has had on my own path, I am equally passionate about dismantling educational inequality and actively partner with non-profits to drive lasting change.You can view my academic publications and research on Google Scholar: & hl=en & oi=ao

Experience

1. Applied Scientist

   Amazon

   Dec 2023 — Present · Bellevue, WA, US

   Production-Scale Geospatial ML: Architected and deployed predictive models for transit time, safety, and driver preferences—serving as core routing inputs for billions of daily last-mile requests- Rigorous Metric Design: Engineered robust measurement frameworks for ambiguous real-world problems, translating complex roadway designs and driver behavioral signals into actionable traffic-safety exposure metrics- Applied VLMs & GenAI: Integrated Vision-Language Models and generative methods to enhance GPS accuracy, audit complex model behaviors, and advance interpretable, trustworthy AI in safety-critical routing- Experimentation & Workflow Innovation: Leveraged GenAI to accelerate applied research workflows, enabling deeper scientific investigation of model outcomes and drastically improving iteration speed- Science Mentorship: Directed an intern project on driver path preference learning, delivering an 8x improvement in preference-detection efficiency over random inspection baselines.

Education

• Carnegie Mellon University School of Computer Science

  Master of Science - MS, Machine Learning

  2019

• Carnegie Mellon University

  Doctor of Philosophy - PhD, Advanced Infrastructure Systems

  2019 — 2023

• Indian Institute of Technology, Madras

  Master of Technology (M.Tech.), Construction Management

  2012 — 2017

• Indian Institute of Technology, Madras

  Bachelor of Technology (B.Tech.), Civil Engineering/Operations Research

  2012 — 2016

• Carnegie Mellon University

  Master of Science, Architecture - Engineering - Construction Management

Skills

• Python
• Scrum
• Project Management
• Autocad
• Web Development
• C++
• C
• R
• Matlab
• Php
• Microsoft Office