Maksym Andriushchenko
Priority directions:
- Risks from automating AI research
- Automating safety and alignment research
- AGI privacy
- Measuring long-horizon agentic capabilities
- New alignment methods
- Science of post-training
Stream overview
I'm interested in all areas of AI safety and alignment, but my priority directions are:
- Risks from automating AI research
- Automating safety and alignment research
- AGI privacy
- Measuring long-horizon agentic capabilities
- New alignment methods
- Science of post-training
Mentors
I am a principal investigator at the ELLIS Institute Tübingen and the Max Planck Institute for Intelligent Systems, where I lead the AI Safety and Alignment group. I also serve as chapter lead for the new edition of the International AI Safety Report chaired by Prof. Yoshua Bengio. I have worked on AI safety with leading organizations in the field (OpenAI, Anthropic, UK AI Safety Institute, Center for AI Safety, Gray Swan AI). I obtained my PhD in machine learning from EPFL in 2024 advised by Prof. Nicolas Flammarion. My PhD thesis was awarded the Patrick Denantes Memorial Prize for the best thesis in the CS department of EPFL and was supported by the Google and Open Phil AI PhD Fellowships.
Perhaps, the best way to understand my research style is to check my website and my most recent papers.
Mentorship style
I usually spend at least 30 min per week in one-one-one meetings with my mentees. We can also discuss longer time slots if necessary. Besides these time slots, I try to be as responsive as possible over Slack (>2 comprehensive responses per day) and read relevant papers between weekly meetings.
Perhaps, the best way to understand my research style is to check my website and my most recent papers.
Representative papers
I'm roughly interested in most directions and papers mentioned in https://internationalaisafetyreport.org/.
Scholars we are looking for
I'm looking for the following skills:
- Prior research experience in a topic related to AI safety (at least one completed project with first-author contribution)
- Independent, self-driven personality
- Strong general computer science background
- Ideally, a good software engineering background
- Familiarity with deep learning frameworks
- Clear communication
No constraints here. I'm fine with both internal (i.e., within MATS) and external collaborators. I can also pair MATS scholars with PhD students in my group, if it's useful.
Project selection
I would prefer to set the overall direction, but I will listen closely to scholars about their preferences within a broad direction. Converging on a particular topic is expected to be a collaborative process.
Community at MATS
MATS Research phase provides scholars with a community of peers.
.webp)
During the Research phase, scholars work out of a shared office, have shared housing, and are supported by a full-time Community Manager.
Working in a community of independent researchers gives scholars easy access to future collaborators, a deeper understanding of other alignment agendas, and a social network in the alignment community.
Previous MATS cohorts included regular lightning talks, scholar-led study groups on mechanistic interpretability and linear algebra, and hackathons. Other impromptu office events included group-jailbreaking Bing chat and exchanging hundreds of anonymous compliment notes. Scholars organized social activities outside of work, including road trips to Yosemite, visits to San Francisco, and joining ACX meetups.