Software Engineer @ DoorDash | AI Safety Researcher | Sailor
I like distributed systems, dev tooling, and AI safety. I've also built a really cool tool that I'm very proud of called Pancake, which is a parallel agent manager, which you can find if you click here. Model Evals are also super interesting and I'm open to discussing new opportunities if you want to reach out! Currently fascinated by interpretability and model evaluations.
For most up-to-date info, check out my website!
https://gabe-mousa.github.io/gabemousa-website/
|
Parallel Agent Manager A fun local node package which allows managing multiple Claude Code or Claude chat sessions. It allows grouping sessions, sending messages across multiple sessions simultaneously, or most interestingly, sessions are allowed to communicate with one-another. This means sending messages back and forth and reading chat sessions of other agents. This creates unique scenarios having sessions spawn sub-agents and being able to live-interact with those sub-agents. |
AI Safety Evaluation Library for Python Building tools to evaluate frontier models for:
Inspired by research from Anthropic's alignment team. |
New Bets
|
gabriel = {
"languages": ["Python", "C#/.NET", "Golang", "C++", "Java"],
"ml_ai": ["TensorFlow", "Scikit-learn", "XGBoost", "Semantic Kernel", "FastMCP"],
"infrastructure": ["Kubernetes", "Docker", "Azure", "GCP", "AWS"],
"research_interests": [
"Mechanistic Interpretability",
"Pragmatic Interpretability",
"AI Alignment Research",
"Chain-of-Thought Faithfulness",
"Sycophancy & Reward Hacking"
]
}
Papers and research that have my attention:
- On the Biology of a Large Language Model — Circuit tracing & attribution graphs in Claude
- Sycophancy to Subterfuge — Reward tampering in language models
- The Case for CoT Unfaithfulness is Overstated — Post-hoc faithfulness analysis
Always looking for paper recommendations in interpretability & alignment!
I'm always open to:
- Chatting about AI safety, interpretability, or alignment research
- Collaborating on open-source AI safety tooling
- Discussing new opportunties if you want to reach out
Reach out: LinkedIn or gab.01@hotmail.com
AI Safety is cool, I'd love to talk if you want to reach out