Company: BrainShells Salary Range: from $7000

We are looking for a strong ML Researcher at the Senior / Staff level for a complex research task at the intersection of probabilistic ML, sequential decision making, and agent behavior modeling. This is not a classic Data Scientist position with tabular data tasks. We need someone who has independently developed or significantly improved ML/mathematical methods, can work with uncertainty, and can experimentally prove why a new approach is better than a strong baseline. Task: There is a stochastic multi-agent environment with partial observability and tens of millions of historical episodes. It is necessary to model the behavior of participants and use these predictions for decision-making under uncertainty and with a limited inference budget. Experience in one or more of the following areas is of particular interest: — RL / sequential decision making / POMDP / world models; — probabilistic & Bayesian ML, latent-variable models, uncertainty estimation; — behavior / opponent modeling, multi-agent systems; — off-policy evaluation, importance sampling, doubly robust estimation, variance reduction; — Monte Carlo / Markov models / stochastic optimization; — algorithmic trading / HFT / pricing / decision systems with real latency constraints. Who we particularly want to see: researchers with a strong mathematical background, experience in their own R&D developments and publications, or strong applied research experience. A PhD will be a plus, but is not mandatory. More important is the ability to break down your project in the format: problem → baseline → your idea → alternatives → experiment → measurable result. Prototyping stack: Python / PyTorch. Data and computational infrastructure are available, the task involves many experiments and a considerable degree of research freedom. Format: remote, contract from 3 months. We prioritize full-time, but are willing to discuss part-time for a strong candidate. If your experience is relevant — send your CV + 2–3 sentences about your most relevant research/ML project: what you improved, what the baseline was, what you did personally, and what result proved the effect.