Skip to content
View danorel's full-sized avatar
πŸ™ƒ
Let's fly to our dreams!
πŸ™ƒ
Let's fly to our dreams!

Block or report danorel

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
danorel/README.md

Typing SVG

Terminal GH GIF

πŸš€ About Me

Hi there! πŸ‘‹

I'm a Machine Learning Engineer who spent two years at Rozetka, Ukraine's largest e-commerce platform, building search, recommendation and auction systems used by millions of people.

Since then I've moved into interpretability research – most recently in the exploration phase of MATS 12.0, in Neel Nanda's mechanistic interpretability stream.

I'm open to research-focused roles in interpretability, LLM evaluation or ML engineering – based in Paris, remote, or happy to relocate, especially to the US.

✨ What drives me

I've always been a builder and a curious mind – from physics papers to tinkering with ML models, there's something genuinely satisfying about understanding how a system works and then pushing it further.

My background is deliberately varied: e-commerce recommendations, auction systems, anomaly detection, a startup, healthcare data. I like jumping between domains because the interesting problems rarely stay in one place – and the best ideas usually come from somewhere unexpected.

For a while I'd been spending more and more of my evenings on AI safety and LLM evaluation – now it's my full-time focus. The more capable models become, the more it matters to understand what's actually happening inside them, not just what they say on benchmarks. What I'm most curious about right now is the gap between what a model knows and what it tells you. It's a young field with a lot of open questions, which is exactly the kind of place I like to be.

And when I'm not poking at activations, I like making things go fast – most recently that meant 1st place at the GPU Mode IRL Hackathon in Paris, with an MoE inference engine doing on 8Γ—H200.

And away from the screen, my newest hobby is building a drone from scratch – and learning to fly it, which turns out to be a whole different kind of debugging.

πŸ§ͺ What I've been working on

bridge-jspace – together with Zach Perlman, asking whether the hidden "bridge" step in two-hop reasoning (Amazon River β†’ Brazil β†’ Portuguese) flows through the part of the representation interpretability tools can read, or somewhere they can't see. The answer depends on where the model gets its answer: recalling a fact from memory leans on what the lens can read, while looking it up in the prompt takes a route the lens doesn't see at all.

secret-elicitation-audit – a model is fine-tuned to secretly know something about the user and deny it, and a benchmark scores how well interpretability tools can pull that secret out. A new probe seemed to beat the published methods, but the whole gain came from the layer it read, not from the probe – and the model grading every method turned out to have a favourite answer of its own. The lesson I keep coming back to: before crediting a method, check what else changed.

Smaller experiments – a refusal-direction replication, chain-of-thought control, and Deceptron, a scheming-detection pipeline built on ground truth you can execute – are in my repositories.

πŸ”¬ What you'll find here

This is my playground for experimenting, learning, and building! You'll find interpretability experiments, ML projects, robotics simulations, physics-inspired projects, and probably some random stuff I'm just curious about. Each project is a learning adventure, and I love sharing that journey.

πŸ’¬ Always happy to connect with fellow builders, learners, and anyone curious about science and technology!

🌐 Connect with Me

To connect with me, you can find my professional profile on LinkedIn or drop me an e-mail. Feel free to reach out for collaborations, opportunities, or just to exchange ideas about interpretability, ML, and technology. Let's build something together!

LinkedIn E-mail

πŸ’» Programming Languages

The languages and libraries I actually work in day to day:

Python PyTorch Hugging Face

πŸ”¬ Research Toolkit

What I reach for when I want to know what's actually going on inside a model:

nnsight vLLM Activation Patching Linear Probes Steering and Ablation LLM Judges

βš™οΈ Infrastructure

What I use to train, serve and ship models in production:

Docker Kubernetes KServe Vespa

Pinned Loading

  1. zach-perlman/bridge-jspace zach-perlman/bridge-jspace Public

    Does the bridge entity of a two-hop question act through the J-space? Probe swaps on open models.

    Python 2

  2. deceptron deceptron Public

    Scheming monitors for coding agents, trained on ground truth you can execute instead of an LLM judge’s opinion: the hard part turned out to be the ground truth itself.

    Python 1

  3. mats.secret-elicitation-audit mats.secret-elicitation-audit Public

    Forked from cywinski/eliciting-secret-knowledge

    Auditing the white-box secret-elicitation benchmark of CywiΕ„ski et al. (2025) on Gemma 2 9B: the gain is the read layer, not the method.

    Python

  4. mats.cot-control mats.cot-control Public

    Banning a word from a model’s chain of thought mostly removes it from the text: inside the model, the ban only shows up in the last layers.

    Python

  5. bomberland bomberland Public

    Bomberland: Reinforcement Learning competition

    TypeScript 1 3

  6. ml.fitness ml.fitness Public

    From 0 to hero! Thanks, Andrej Karpathy

    Jupyter Notebook