GitHub
Hey, nice to meet you! I’m Jason (aka David Kim), human, life-long learner, and AI engineer!
I work on AI security and agentic evals as a Hackerone Red Teamer evaluating foundational LLM architectures and classifiers for Anthropic while conducting research on context engineering, multi-agent orchestration for production, and hallucination prevention.
As part of my Prime Intellect RL Residency, I’m also conducting research into automating the full hypothesis generation + validation pipeline with reinforcement learning and universal verifiers (LLM-as-judge).
Outside of research, I enjoy thinking about the future, science fiction, and visualizing complex systems in ways that make them more human-understandable. I'm always curious about how we can use novel methods, not just scale, to solve problems more ethically and intelligently. Thanks for visiting and feel free to explore my work or connect. I believe there's a lot to learn ahead.
– Jason Tang
Open Source Contributions
Quant Lab | Cognitive Tools | Reinforcement Learning 101 | TransformerCircuits | NVIDIA Universal Deep Research | AISecForge | Context Engineering | kernels
Google ML Implementations:
Linear Regression | Binary Classification | Logistic Regression
Research
- Model Research Instruments: An Exploratory Behavioral Study for Automating Scaffolded Model Hypothesis Generation
- critic: An Exploratory Behavioral Prevention Study on AI Sycophancy
Postgraduate Education
Google: Machine Learning Education
Stanford: CS 229 (Machine Learning), CS 230 (Deep Learning), CS 231N (Computer Vision), CS 336 (Language Modeling from Scratch)
Harvard: CS 2881 (AI Safety)
UC Berkeley: CS 285 (Deep Reinforcement Learning)
ARENA: AI Safety Bootcamp (Fundamentals, Transformer Interpretability, RL, LLM Evaluations)