Who we are
Polished Snow is an AI lab building energy-based reasoning models for critical systems. Power grids, industrial infrastructure, the technology the world runs on.
Our model, Tars, has scored top on multiple formal verification benchmarks. The reason it performs is architectural. Most AI models are built to generate a plausible answer. Tars is built to know when an answer is wrong before it becomes one. It scores, refines and verifies a plan against the full system until it holds. That difference matters enormously when the cost of being wrong is measured in billions.
We are a small, funded team and we are just getting started.
What we're looking for
A Research Engineer who thinks from first principles and builds things that work outside the lab. You care about why models behave the way they do, and you care that they behave correctly when it actually matters. If you have worked on generative models, probabilistic systems, formal verification, or large-scale ML and you are drawn to the hardest reliability problems in AI, this is the role.
What you'll do
Push what Tars and energy-based reasoning can do beyond where current autoregressive and diffusion approaches stop. New architectures, new training paradigms, new ways of representing constraints, correctness and system state.
Take ideas from hypothesis through experiment to working prototype, and then into production systems running inside critical infrastructure.
Design and run experiments, training pipelines and inference systems that make verified reasoning possible at scale.
Work across research and engineering without a hard line between them. What you build will be recognized by both sides as theirs.
Help define the scientific direction of the lab. What we work on next, how we evaluate it, and where we publish when it matters.
Who you are
Deeply technical, with a track record in generative modeling, probabilistic inference, large-scale ML systems or adjacent areas. You have either originated work that moved the field, shipped systems at scale, or both.
Comfortable working from first principles. You can define a problem, design the experiments and build the minimal system that proves what is possible.
You care how and why models work, and you care that they actually work when the stakes are real.
You have built in areas without precedent and created the tools and frameworks as you went.
You have a real view on where reasoning in AI needs to go next. Bigger model, more compute is not an answer that lands here.
You thrive in small, focused teams that value autonomy, speed and tight collaboration over process.
You think AI for critical systems is one of the most consequential problems in the field right now and you want to be early to solving it.
Compensation Range: $140K - $220K
