Jeffrey Zhou

Strengthening various integrity/identity systems that help mitigate fraud, attacks, and scaled abuse against ChatGPT and Codex while reducing authentication/verification friction for legitimate users to boost key growth metrics.

Accelerating large-scale Autopilot/Optimus distributed training. Built core features and infra for Tesla's ML platform to visualize high-loss samples at train-time and analyze RL reward behavior. Contributed to Full Self Driving (FSD).

Built PoCs employing hyperscale embedding gen/retrieval and agentic AI. Implemented inference optimizations on Arm-based devices including quantization and KV-cache offloading for Arm's next-gen AGI CPU.

Work with DARPA and MIT researchers to build RESPECT, a Responsible AI benchmarking + evaluation framework for testing autonomous military systems using LLM-guided adaptive scenario generation. See DARMA.mil and Saab.com

1. Student Researcher

   Sep–Dec 2024

   Developing and testing a Python package framework for improving conversational agent performance using Finite State Machine-guided LLMs.

2. Student Researcher (SURF)

   Jun–Sep 2024

   Implemented MLP models and time-series imaging for NASA's ACE mission to classify decay phases in flux data for solar energetic particle events.

3. Student Researcher (FSRI)

   Jun–Sep 2023

   Built experimental modular rocket engine injector test stand for hydro-patternization analysis. Conducted fluid/thermal simulations on various injector designs.