Work with me

Status:

Recruiting PhD students for Fall 2026.
Open to new collaborations.

Research Focus

I’m broadly interested in the design, learning, and specification of goals, rewards, and abstractions for intelligent agents, including reinforcement learning and language model agents. How can we rigorously specify or determine what we want our agents to do?

Advising Philosophy

My primary goal is for you to become an independent researcher capable of driving your own agenda.

The Goal: For you to develop the “research taste” to identify important problems as well as the technical agency to solve them.
The Path: I view advising as a partnership. Early on, we will work closely to scope projects and build strong foundations. As you develop, I will step back to support your autonomy.
The Opportunity: As an incoming Assistant Professor, I am starting a new research group. This offers a unique opportunity to play a significant role in defining our early research culture.

Technical Fit

I am looking for students with strong mathematical intuition and an interest in alignment, multi-objective problems, sequential decision making, reinforcement learning, and/or language model agents.

I do not expect you to be an expert in everything. Instead, I look for students who have developed deep strength in one main area, or who have the foundational strength to acquire new skills rapidly.

Examples of profiles that would fit well:

The Axiomatic Thinker: You enjoy thinking about the properties of representative objects (e.g., $ℝ 𝑁$ as a metric space) and formalizing vague concepts. You might have an interest in decision theory, representation theorems, or axiomatic alignment work.
The Pragmatic Theorist: You have a rigorous background (e.g., Math, Physics, Theoretical CS) but are eager to apply it to practical problems. You want to use your technical toolkit to build and understand agents that function in the real world.
The Agent Architect: You are interested in generalist agents and reasoning. You follow the latest developments in language models and are interested in how they can be used for planning, exploration, and sequential decision-making.
The Self-Driven Learner: You may not fit perfectly into the boxes above, but you are a rigorous autodidact. You have a track record of picking up complex technical areas (e.g., a new field of math or a new codebase) very quickly on your own.

A note on ‘fit’: Research is about your trajectory, not just your current state. If you have a strong technical foundation and are driven to define your own research path, please apply!

Undergraduates & Master’s Students

I am open to working with motivated UMich students who demonstrate agency.

If you have a specific hypothesis you want to test or a research direction you have already started exploring, please reach out.
I am much more likely to reply to a student proposing a specific idea than one asking generally to “join the lab.”

How to Apply

1. PhD Applicants (Fall 2026)

You must apply through the University of Michigan CSE PhD Program, and list Silviu Pitis in your application.

2. Contact Protocol

I welcome emails from prospective students, but please follow this format to ensure your email is seen:

Subject Line: Please include prefix [F26-Applicant].
Content: Please keep it brief, highlight your key interests and relevant experience, and include your CV.

Note: I get a lot of emails, and can only respond to a few. If you feel you are an exceptionally strong fit, please follow-up after a week or two if I haven’t responded.