Igor Krawczuk
M.Sc. (Ph.D. cand.), LIONS@EPFL
igor.krawczuk (at)


  • I am currently a researcher and Ph. D. student at the LIONS laboratory at EPFL. My current research focuses on generative models and reinforcement learning for combinatorial domains, in particular their application to integrated circuit design.
  • To see what this means I do concretely, check my favourite works, my CV a summary of my high level research directions and my selected side projects
  • As my Ph.D. nears its end, I am now looking for suitable applied ML/ML research or leadership roles in tech/finance firms or private research entities. If you are looking to fill such a role based in Switzerland, please check below whether my profile fits and reach out :-) Also open to consulting and other types of collaborations, so if you want to chat book a slot at

Details about me

Since this is the sales pitch, let me try to overcome the academic impostor syndrome: I am a T-shaped person who has done work on almost all aspects of computing, machine learning and optimization.

Originally trained as an electrical and computer engineer, I am currently a researcher and terminal Ph.D. student at the LIONS laboratory at EPFL. My doctoral studies started with neuromorphic computing hardware - trying to build brain inspired ML accelerators - and slowly drifted into applied ML and ML/optimization theory.

This has given me the chance to work with the “full stack” of ML research and engineering. During my Ph.D. I have done hardware design, systems programming with C,C++ and Rust, written my own CUDA kernels as well as large scale ML and RL systems in tensorflow, pytorch and jax. I also supervised over 20 students and interns and worked with half a dozen industrial partners. Finally in the last decade I’ve not only been been a Ph.D. student, but also the co-founder of 3 startups, a software and ML prototyping consultant, the sys-admin of our lab, a contributor in lobbying for improved AI governance, and a board-member of various association and NGO boards.

My interests

While my current research focuses on sample efficient distributional modeling and reinforcement learning methods on combinatorial domains like sets and graphs, I got into ML wanting to understand decision making processes in adaptive systems.

Together with a politically conscious socialization in the german hacking scene (particular CCC influenced online spaces) this has developed into studying the intersection of computer science and politics and economics , although my lens is of course mainly shaped my engineering and ML background.

As such you can split my research into a technical track and into a humanities (polisci/economics/governance) track.

On the technical side I am interested in using ML methods to study and improve corporate, economic and political decision making processes1. I have also been drawn to more applied industrial problems like medical imaging or industrial control systems, where I care about ensuring the correctness, robustness and fairness of the deployed systems.

On the humanities side I am interesting in the interaction of specific algorithmic systems and their biases with humans and society at large. In particular, I want to contribute to mechanisms of accountability that ensure the benefits of AI systems are not monopolized or abused 2 and improve our understanding of governance structures in general 3.

Somewhere in between is the subject of algorithmic mechanism design and game theoretical analysis of existing polito-economic systems, for which I gave a rough outline for a research plan in a talk aimed at a non-technical audience at assoziation E.

Looking for new opportunities

As I wrap up my Ph.D. I am looking for new opportunities that allow me to pursue these interests in a research, applied research or leadership role that allows me to remain in Switzerland physically (this is a hard constraint). In particular, I am interested in working with private enterprises (tech firms, financial firms, startups and foundations or private labs) towards the end of 2023.

A non-exhaustive list of roles for which I think I might be a good fit is given below, but I am also open for consulting work, or out of the box ideas. Generally, I’ve found that the most interesting opportunities are those you didn’t expect, so please don’t hesitate reaching out! (Email at the top).

Roles which I consider myself a good fit for:

  • Researcher, senior ML engineer or a technical leadership role in established tech companies or startups interested in applying rigorous ML and optimization techniques to real world problems. In particular problems involving:
    • combinatorial optimization
    • formal verification/guarantees
    • game theoretic problems like robustness, mechanism design (think auctions) etc.
  • Researcher, research manager in private research entities like foundations or thinktanks which would allow me to work on the above problems, or on the intersection between politics, economics and AI. This can include practical work like PoC technologies to improve governance and policy making, pure research or something in between.
  • Researcher, Quantitative researcher or ML engineer for financial institutions or trading houses who can make use of my game theory and combinatorial optimization academic background as well as my software engineering skills to develop new ways of managing risks.
  • Entrepreneur in residence, innovation manager or ecosystem designer for established firms or institutions who want to open themselves up to open source and community based innovation to improve their agility and innovative power

  1. An example for governance would be using natural language processing to quantify the consistency between the proclaimed values of corporate and political entities and their behavior in markets/the political process, while an industrial application would be portfolio optimization, risk managing, forecasting or autonomous industrial online-bots and physical drones. ↩︎

  2. See e.g. this Science publication I contributed to in the context of AI governance ↩︎

  3. See e.g. publication on a complexity theory based view on policy making ↩︎


  • Machine Learning on and with graphs
  • Optimization theory and applications
  • Game theory
  • Complex adaptive systems
  • Algorithmic Mechanism Design
  • AI/ML governance
  • Digital Humanities
  • (getting into) Category theory and formal verification


Ph.D. cand. Electrical Engineering
M.Sc. Electrical and Computer Engineering
B.Sc. Electrical and Computer Engineering

Selected Publications

Please see my google scholar for the complete list. * and other symbols denote equal contributions.
DiGress: Discrete Denoising diffusion for graph generation, 2022, preprint, accepted to ICLR 2023
, *Clement Vignac , *Igor Krawczuk , Antoine Siraudin , Bohan Wang , Volkan Cevher , Pascal Frossard
Optimization Theory, Economics and the Real World, 2022, Summary of a nontechnical talk about what I perceive as important gaps in our formal economic and political toolkit from an optimization theoretic perspective. Given at assoziation E in the context of a colloquium on Utopias and Dystopias
Igor Krawczuk
Maxime Stauffer , Isaak Mengesha , Konrad Seifert , Igor Krawczuk , Jens Fischer , Giovanna Di Marzo Serugendo
Shahar Avin , Haydn Belfield , Miles Brundage , Gretchen Krueger , Jasmine Wang , Adrian Weller , Markus Anderljung , Igor Krawczuk , David Krueger , Jonathan Lebensold , Tegan Maharaj , Noa Zilberman
GG-GAN: A Geometric Graph Generative Adversarial Network, 2020, PrePrint(submitted to ICLR2021)
*Igor Krawczuk , *Pedro Abranches , *Andreas Loukas , Volkan Cevher

Some non-academic projects

Time Scarfe interviewed my friend Carla Cremer and me about our perspectives on X-Risk, Governance in EA and AI risk. I wish we had had time to go into my points about all reasons I don’t believe in AGI risk, but I am still glad we could make our points in such a nice format.

Although I am NOT an EA for a various reasons, the Swiss EA communities’ focus on AI and governance mean I keep hanging out with them. I’ve served as a board member of the EA Geneva association, have tried to bring a diversity of philosophies and angles of criticism into the community as a part of their facilitator program and sporadically give talks about AI risk.

I’m a board member of the alumni association of the Manage And More entrepreneurial scholarship, serve as a Mentor to young Entrepeneurs (e.g. as part of the 2022 ESADE eWorks program) and offer online ‘office hours’ for young founders to critique their business and technology stack. I also occasionally give talks at events, e.g. about blockchain technology at BSL

Appeared as one of the experts on a special program on AI and algorithmic governance organized by the Swiss German national TV station SRF