Autmos designs and validates autonomous systems — spanning aerial robotics, perception, control, and learning-enabled architectures — with an emphasis on engineering realism, safety, and reliable operation under real-world constraints.
Full-stack autonomy for UAV systems operating in constrained, GPS-denied, or adversarial environments — from guidance and control to mission-level decision-making.
Sensor fusion, mapping, and representation learning for reliable navigation and situational awareness under degraded or denied sensing conditions.
Physics-aware control design, collision avoidance, and guidance architectures developed with rigorous attention to closed-loop stability and failure modes.
Integration of AI and machine learning components within safety-critical pipelines, maintaining interpretability and compatibility with real-world operational demands.
Full-stack validation workflows bridging software-in-the-loop and hardware-in-the-loop testing through to field experimentation and prototype deployment.
Embedded system design, rapid prototyping, and hardware integration — with consistent attention to runtime, power, and sensing constraints.
Work begins by clarifying the structural properties of the problem — system dynamics, uncertainty sources, observability limits, and computational constraints — before any algorithm or model class is chosen. Solutions evolve through iterative refinement across simulation, hardware, and field validation.
The emphasis throughout is on coherent system architecture rather than isolated component performance. Perception, inference, control, and dynamics are designed to interact reliably — not optimised in isolation.
Dominant constraints and failure modes are identified first. Components are designed in the context of the full closed-loop system, not in isolation.
Physics-based priors are balanced with data-driven methods where appropriate. Neither is applied dogmatically.
Validation pipelines are structured to surface discrepancies early — before they propagate to hardware and field trials.
Solutions are developed under real hardware, sensing, and runtime constraints from the outset — not retrofitted to them at the end.
Developed full estimation pipeline — sensor fusion, mapping, and localisation — validated across a simulation-to-field workflow including indoor and GPS-denied outdoor environments.
Designed real-time avoidance and path-following system with formal analysis of safety margins and hardware-constrained runtime performance.
Built high-fidelity simulation and test infrastructure to accelerate iteration cycles and systematically characterise failure modes prior to field deployment.
Integrated deep learning perception components into a closed-loop autonomous system, with explicit attention to stability, distributional robustness, and interpretability requirements.
Led parallel prototyping and research program spanning perception, control, and embedded system design. Defined technical direction, managed cross-functional teams, and aligned execution with safety and deployment constraints.
Autmos is an independent practice in autonomy, robotics, and AI systems engineering — built at the intersection of aerial systems and the broader challenge of making autonomous intelligence reliable in the physical world.
The work spans control systems, estimation, perception, simulation, and learning-enabled design, with a consistent focus on bridging theoretical models and real-world deployment constraints. Engagements range from early-stage architecture definition through to full prototype development and field validation.
Based in Toronto. Working with teams across research, defence, and advanced industry.