Capabilities
Research and integration across the entire path from observations to decisions and physical actions.
Recurrent and attention-based architectures; convolutional sequence models; Transformers; tree-based baselines; probabilistic prediction; feature selection and target design.
Custom environments, action spaces, experience replay, exploration strategies, and reward design. Current robotics work explores pretrained priors, supervised adapters, and parameter-efficient policy refinement.
Simulator alignment, domain randomization, perturbation testing, and independent validation paths using IsaacLab and MuJoCo.
Model export, ONNX inference, preprocessing, history buffers, scheduling, action scaling, and control-loop synchronization.
Working together
Engagement scope is shaped around the research question, available data, deployment constraints, and evidence needed to evaluate the result.
Define the objective, baseline, assumptions, and success criteria.
Develop the environment, representation, model, or integration layer.
Evaluate failure modes, distribution shift, and runtime consistency.
Walk through the methodology, findings, limitations, and next decisions.
Start a conversation
Discuss a research collaboration, explore a training pipeline, or request a walkthrough of our work.
Discuss your project