Aman Priyanshu
Principal AI Research Scientist
San Francisco, California, United States · 1 connectionNot on Unlinked · Imported profile. Is this you? Claim it
Is this about you? Request removal
About
Hi, I'm Aman! I'm an AI Researcher at Foundation AI building foundation models for security applications. My work spans the complete training pipeline - from pre-training LLMs over billions of tokens with distributed training and model souping, to post-training with Group Relative Policy Optimization (GRPO), Direct Preference Optimization (DPO), and Anchored Preference Optimization (APO). I develop synthetic data generation methods for knowledge acquisition, structured responses, and tool usage that enable agentic orchestration. I also build custom reinforcement learning environments for security reasoning and vulnerability discovery. Previously in AI safety and security, I uncovered critical vulnerabilities in foundation models at Robust Intelligence, including research that achieved a 99.8% bypass rate on Meta's PromptGuard system, which garnered media attention in The Register and SC Media. My various efforts in AI safety led to invitations to join OpenAI's Red Teaming Network and Anthropic's Model Safety Bug Bounty Program. With a Masters in Privacy Engineering from Carnegie Mellon University, I've collaborated with Professor Norman Sadeh on LLMs and cybersecurity research, and
Experience
Jan 2024 to May 2024
Apr 2023 to Jul 2023
Jul 2022 to Apr 2023
Jan 2020 to Feb 2020
Education
Masters Privacy engineering
Bachelor of Technology - BTech Information Technology