Nikil Ravi
  • Thoughts
  • Notes
  • Stuff I found interesting:
  • Work
Home

Work

  • Valkyrie: scalable, cloud-native infrastructure for evaluating AI agents across benchmarks.
  • FormalProofBench: benchmark for evaluating whether frontier models can write graduate-level Lean proofs that pass formal verification.
  • Marin: open lab for building fully open foundation models with transparent experiments, data, and training code.
  • REAL: benchmark for autonomous agents on deterministic simulations of real websites.
  • FAIR principles for AI models: Scientific Data paper proposing practical FAIR principles for AI models, with an application to accelerated high-energy diffraction microscopy.
  • TensorRT tutorial: tutorial on optimizing deep learning inference with NVIDIA TensorRT; video.
  • Meta Sparse: estimating task relationships using subnetwork similarity; report.