Paul Kassianik
Principal AI Security Researcher at Foundation AI
San Francisco, California, United States
Summary
🤩
Rockstar🎓
Top SchoolPaul Kassianik is a Principal AI Security Researcher with 11 years of experience building scalable pretraining pipelines, production-grade guardrails, and adversarial testing frameworks for enterprise and startup environments. He led development of Foundation-Sec-8B, an 8B-parameter cybersecurity LLM, and architected AI-Firewall defenses and proactive threat taxonomies now used in live security workflows. His published work (NeurIPS, ICML) on automated red-teaming and jailbreak reasoning has set benchmarks for adversarial evaluation and driven hundreds of citations, while his synthetic jailbreak frameworks cut time-to-product integration to days. Previously at Salesforce he delivered a code auto-complete system and the BigIssues bug dataset, improving developer productivity by ~30% for over a thousand engineers. Equally comfortable in research and production, Paul contributes to open-source ML tooling (notably backend work on the juice ML engine) and briefs CISOs and policymakers on AI risk mitigation. Based in San Francisco with a BA in Applied Mathematics from UC Berkeley, he combines rigorous analysis with pragmatic engineering to move safety research into operational impact.
11 years of coding experience
6 years of employment as a software developer
Bachelor of Arts - BA, Applied Mathematics, Bachelor of Arts - BA, Applied Mathematics at University of California, Berkeley
English, Russian
Stackoverflow
2,203reputation
180kreached
60answers
58questions
Github Skills (17)
application-framework10
app-framework10
machine-learning10
web-framework10
rust10
linear-algebra10
cuda9
opencl8
testing7
while-loop6
android6
android-sensors6
jquery6
multithreading6
java6
Programming languages (13)
Github contributions (5)
fff-rs/juice
Dec 2019 - Mar 2020
The Hacker's Machine Learning Engine
Role in this project:
Back-end Developer
Back-end DeveloperContributions:6 reviews, 25 commits, 8 PRs in 3 months
Contributions summary:Paul primarily contributed to the implementation and improvement of the `juice` machine learning engine. Their work involved adding features to the `linear_layer`, addressing compilation issues by incorporating the `Copy<f32>` trait, and fixing bias calculations and matrix shapes. They also added regression tests for the linear layer, demonstrating a focus on testing and verification.
cudacoasterpythonjuiceagnostic
paulkass/stunning-meme
Mar 2019 - Oct 2023
Contributions:60 pushes in 4 years 7 months
Find and Hire Top DevelopersWe’ve analyzed the programming source code of over 60 million software developers on GitHub and scored them by 50,000 skills. Sign-up on Prog,AI to search for software developers.