2025
In-Context Learning Without Copying
arXiv preprint
Read paper →Hello, I'm
Adam Belfki
I'm currently building infrastructure for studying the internal mechanisms of large AI models at NDIF and doing AI interpretability research at Baulab.
About
Building the future of AI, responsibly.
I'm a research engineer with deep experience in distributed systems and neural network architectures. Currently, I'm building a robotics integration platform to simulate, test, and deploy multi-agent systems.
My research interests lie at the intersection of AI safety and interpretability—understanding how neural networks work internally to ensure they behave reliably and align with human values.
Current Focus
Robotics & Multi-Agent Systems
Research
Mechanistic Interpretability
Languages
Arabic • English • French
Publications
Selected Work
2025
MIB: A Mechanistic Interpretability Benchmark
ICML 2025
Read paper →2024
NNsight and NDIF: Democratizing Access to Open-Weight Foundation Model Internals
ICLR 2025
Read paper →2024
Dyna-5G: A Dynamic, Flexible, and Self-Organizing 5G Network for M2M Ecosystems
arXiv preprint
Read paper →2023
Analyzing the Impact of GNSS Spoofing on the Formation of Unmanned Vehicles Swarms
ION GNSS+ 2023
Read paper →Research Interests
What I'm exploring
Mechanistic Interpretability
Reverse-engineering neural networks to understand the algorithms and representations they learn internally.
AI Safety & Robustness
Developing methods to ensure AI systems are reliable, predictable, and aligned with human intentions.
Multi-Agent Robotics
Building platforms for simulating, testing, and deploying coordinated robotic systems.
Distributed Systems
Architecting scalable, fault-tolerant infrastructure for large-scale AI and data processing.
Get in Touch
Let's chat
I'm always interested in discussing research, collaborations, or just connecting with like-minded people.