Keshav Pradeep

I’m currently a PhD student at Cornell University, advised by Dr. Rachee Singh. My research is supported by the NSF Graduate Research Fellowship Program.

Before starting my PhD, I was an engineer at NVIDIA working on scaling Agentic AI, including agent skills, harnesses, multi-agent systems, knowledge graphs, long-term memory, guardrails, and enterprise-scale token savings.

Before that, I completed my undergrad in CS & Math at the University of Maryland, where I researched collective communication in deep learning workloads and network-induced variability in exascale GPU-accelerated supercomputers under Dr. Abhinav Bhatele in the Parallel Software and Systems Group.

Conference Papers

• The Big Send-off: Scalable and Performant Collectives for Deep Learning
  Siddharth Singh, Keshav Pradeep, Mahua Singh, Cunyang Wei, Abhinav Bhatele
  IPDPS 2026 · doi · pdf · website

• The Case of the Elusive Application Performance on Production GPU Supercomputers
  Cunyang Wei, Keshav Pradeep, Abhinav Bhatele
  IPDPS 2026 · doi · pdf · website

Posters

• Centralized AI Resource Platform (Eco)System: Making Skills from idea to trusted use
  Mohit Gupta, Yogesh Dangi, Keshav Pradeep, Siddharth Itagi, Narendran Raghavan, Yashraj Basavaraj Patil, Wei Luo, Subodh Prabhu, Christopher Kevin, Amanda Burgmeier, Jean-Francois Puget, Pranita Maske, Moshe Abramovitch, Daniel Major, Alban Douillet, Ryan E. Wong
  AI Agents Workshop at KDD 2026

• SkillSpector: A Pre-Publication Security Control for Agent Skills
  Nir Paz*, Keshav Pradeep*, Narendran Raghavan, Ashley Nikirk, Mohit Gupta (*equal contribution)
  Agent Skills Workshop at CAIS 2026 · poster · openreview · repo

• Optimizing Collectives with Large Payloads on GPU-based Supercomputers
  Siddharth Singh, Mahua Singh, Keshav Pradeep, Abhinav Bhatele
  SC 2025 · poster · abstract

• SC 2025 Best Poster Finalist: Unmasking Performance Variability in GPU Codes on Production Supercomputers
  Cunyang Wei, Keshav Pradeep, Abhinav Bhatele
  SC 2025 · poster · abstract

• Unmasking Performance Variability in GPU Codes on Production Supercomputers
  Cunyang Wei, Keshav Pradeep, Abhinav Bhatele
  MUG 2025 · poster

Open Source

• NVIDIA/SkillSpector
  Security scanner for AI agent skills that detects vulnerabilities, malicious patterns, and security risks before installation.
  SkillSpector was briefly trending on GitHub with 15k+ stars ✨

• NVIDIA/SkillEvaluator
  Multi-tier framework for evaluating AI agent skills with quality gates, semantic overlap detection, synthetic evaluation dataset generation, and live agent evaluation.