LLM TRAINING · INTERACTIVE TUTORIAL
Tinker in Practice
An interactive guide to using Tinker, from your first training loop to fine-tuning, reinforcement learning, and deployment. Learn through visual explanations and hands-on modules.
RESEARCHER & ENGINEER
Chris Kambimbi.
Understanding AI. Making it safer.
I’m a Research Engineer at LibrAI, investigating vulnerabilities in large language models. My work centers on AI safety, red teaming, and alignment.
Experience & education
Experience
2026–Present
Research Engineer
LibrAI
2025
AI Safety Student Researcher
LibrAI
2023–2026
AI Safety Researcher
Fudan University · FVL Lab
2021–2023
AI Operations & Community Specialist
Dazzle AI, CivitAI, ET/Larix
Education
2023–2026
M.S. in AI & Software Engineering
Fudan University
2019–2023
B.S. in Electronic Information Engineering
Tongji University
Languages
English
Professional working fluency
Chinese
Professional working fluency
Blog
Notes, research & reflectionsResearch
Sep 16, 2026Forty Open Problems in AI SafetyA 2026 update to Concrete Problems in AI Safety: forty open research problems, and the funders committed to them.AI Safety / Open Problems / Research · 26 min readSep 15, 2026Continuous Red Teaming for AI AgentsWhy safety testing must operate at the speed of AI agents.AI Safety / Red Teaming / Agents · 26 min readDec 4, 2025DeepSeek V3.2: Reasoning-First Models Built for AgentsHow an open-source release closed the gap with frontier models: sparse attention, large-scale RL post-training, and 1,827 synthesized agentic task environments.AI / LLM / Agents / Research · 36 min readProjects
A few things I’ve builtAI SAFETY · RESEARCH TOOL
MedLLM Attack Taxonomy
An interactive taxonomy of adversarial attacks on medical large language models. Mapping vulnerabilities to better understand their risks.
Publications
Research outputResearch on Automated Jailbreak Safety Evaluation for Large Language Models
Master's thesis · Fudan University