Michael Kamp

9/
If you’re serious about science, and about doing it sustainably, Germany might be the place.
Get in touch.
Ask questions.
Apply.
Let’s build something worth building.
(End 🧵)
(13/13)

8/
We don’t promise Google salaries.
But we offer something different:
Scientific freedom.
Stability.
Time to think deeply.
And colleagues who care more about ideas than metrics.
(12/13)

7/
Germany isn't just Berlin.There are strong groups everywhere.Decentralization is a feature, not a bug.
In many places, you’ll have brilliant colleagues and more room to shine.
(11/13)

We have a fantastic collaboration with the Institute for AI in Medicine in Essen, which means our work actually reaches the clinic. If you're passionate about blending cutting-edge ML research with meaningful real-world impact, we might just be the perfect place for you.
(10/13)

and we also delve into federated learning and privacy. But we don't stop at theory; we bring it all into healthcare, working on real-world applications like predicting chronic disease and developing AI that impact patient care.
(9/13)

6/
Speaking of which…
Our group at the Lamarr Institute focuses on Trustworthy AI for Healthcare. We're a bit of a unicorn: we love diving into the theory, working on things like generalization bounds and robustness through loss surface analysis,
(8/13)

5/
Big names? We have them.
– Max Planck
– Helmholtz
– AI centers of excellence
And yes, some groups do groundbreaking stuff without a famous label.
(7/13)

4/
Strong research without toxic pressure.
You want to write good papers, not grind yourself to burnout.
In Germany, you get time to think - and breathe.
The system isn't perfect, but it’s sane.
(6/13)

without the constant pressure of chasing the next big grant. It’s a place where genuine curiosity and collaboration often take center stage, allowing you to grow not just as a researcher, but as a thinker.
(5/13)

3/
Let’s dive a bit deeper into what sets Germany apart: beyond the excellent funding and paid positions, there's a unique culture of scientific freedom. Here, researchers are encouraged to explore bold, innovative ideas
(4/13)

2/
No tuition fees.
Most PhD and postdoc positions are paid jobs.
That’s right:
– Monthly salary
– Full health insurance
– Retirement and unemployment insurance
– Vacation
– Parental leave
Academic precarity? Not here.
(3/13)

1/
Germany doesn't do hype well.
But we do substance.
f you're into real science, good working conditions, and long-term impact — read on.
(2/13)

Thinking about doing a PhD or Postdoc in ML?Don’t just look at Stanford, Oxford, or ETH.Look at Germany.
Yes, Germany.
(Thread 🧵)
(1/13)

You matter.

Unless you multiply yourself by the speed of light squared.

Then you Energy.

After a long grey German winter finally a sunny evening by the Rhein - sunset, wine, and the gentle chaos of people learning to relax again. Bonn does it right.

At the ETIM '25 they had an artist produce graphical abstracts of all talks. I really love it! Especially the data-hungry model lurking in front of the data fridge... Thanks #IKIM for organizing this event.

🔗 If you or someone you know is interested, let’s build the future of Trustworthy AI together! 🚀

#MachineLearning #PhD #AI #FederatedLearning #Causality #DeepLearning #GPUComputing

trustworthyml.de

🧩 Want to apply?
📩 Send your letter & CV to michael.kamp(at)http://uk-essen.de
📌 Subject: "Application PhD Position Trustworthy Machine Learning X", where
X = G(5) with G(0) = 0, G(1) = 1, and G(n) = G(n-1) + 2*G(n-2).
(7/7)

📍 Why Germany?

- Massive state investments in AI research
- EU Blue Card visa makes relocation easy for non-EU citizens
- Unlike the US, AI research careers here are stable & well-funded
(6/7)

🚀 Unique research opportunities

Physics & AI: One of TU Dortmund's ML papers was cited in the 2024 Nobel Prize in Physics
Medical AI: Work directly with doctors & real patient data—AI that actually improves lives
(5/7)

🌍 Work in one of Europe’s most vibrant AI research environments!

The Lamarr Institute is one of six AI excellence centers in Germany (TU Dortmund & Uni Bonn)
The Institute for AI in Medicine (IKIM) at University Hospital Essen is Europe’s leading center for medical AI
(4/7)

🔥 Why this position?

Fully funded (~58,000€/year, ~3,000€/month after taxes)
Top-tier health insurance, retirement, and unemployment benefits
World-class compute resources:
⚡ Large GPU cluster @ Lamarr and KITE GPU Cluster @ IKIM

(3/7)

💡 Who should apply?
🔢 Excellent mathematical skills (probability, optimization, linear algebra)
🧠 Strong ML background (deep learning, PyTorch, state-of-the-art architectures)
🎯 Passion for theoretical AI with real-world applications
(2/7)

PhD Position in Trustworthy AI at the Lamarr Institute (TU Dortmund, Germany)!

I’m looking for outstanding PhD student working on:
✅ Federated & Multi-Agent Learning
✅ Theory of Deep Learning
✅ Causality and Causal Representations

(1/7)

7/ For all the nitty-gritty details, check out the full paper ("Little is Enough: Boosting Privacy by Sharing Only Hard Labels in Federated Semi-Supervised Learning"): arxiv.org/pdf/2310.05696 📖 #AI #MachineLearning #ResearchPublication

6/ What about interpretable models? They are often more trustworthy but often cannot be trained by federated learning. 🌲 FedCT doesn't discriminate against interpretable models. Works with decision trees, XGBoost, etc. Quality similar to centralized training. #InterpretableAI

5/💡Can we go even one step further? Federated Co-Training (FedCT) forms a consensus over local hard labels as a pseudo-labeling for the public dataset to augment local training data. 📊 FedCT has similar test accuracy as FL and DD with near-optimal privacy. #PrivacyProtection

4/ 🤝 Can we share less data while maintaining model performance?🌐In many fields, vast unlabeled datasets are publicly available. Think healthcare databases. 📈 We can leverage this to share information between clients: e.g., Distributed Distillation (DD) uses co-regularization.

3/ 🔒 Differential privacy offers a theoretical solution, but can it be practically applied? Our findings show that it improves privacy, but not too much, at the cost of model quality. Moreover, practical guarantees are poor (ε = 145, δ = 10−5, arxiv.org/pdf/2210.03843).

2/🏥 Data is often distributed and cannot be pooled - think healthcare. 🚫 Federated Learning may seem like the answer, but our practical results might surprise you: vulnerability against membership inference attacks is high (VUL is probability of successful attack). #HealthTech