Localyte

AI-enhanced inertial sensing for reliable navigation when external references fail.

Localyte is the deep-tech startup I co-founded to make motion sensing and navigation more reliable through software. It sits at the intersection of robotics, sensor fusion, embedded AI, and real-world industrial validation.

The idea is simple: many systems already have inertial sensors, but those sensors drift over time. Localyte improves the quality of the motion data already available, without forcing a hardware redesign.

Drones, robots, vehicles, satellites, and industrial measurement systems rely on IMUs to estimate how they move. IMUs are fast and compact, but their errors accumulate. When GNSS, vision, odometry, or other external references become weak or unavailable, drift becomes a serious limitation.

Localyte works as a software layer between the IMU and the existing navigation or control stack. It analyzes inertial signals in real time, recognizes motion patterns, and applies dynamic corrections to reduce drift, bias, and non-linear errors.

• Software-first

  Improve the sensors already on the platform instead of adding more hardware.

• Compatible by design

  Provide better inertial data to existing estimators, autopilots, and navigation pipelines.

• Physics-aware AI

  Use learning where it helps, while respecting the dynamics of the physical system.

• Industrial testing and NDT

  Segment operator motion and associate measurements with the correct position on a test surface, reducing manual post-processing and making workflows more repeatable.

• UAVs

  Improve navigation continuity in GNSS-challenged missions by feeding more reliable inertial data to the existing autopilot or estimator.

• Space and small satellites

  Reduce gyro drift during periods where absolute attitude updates are unavailable, supporting better pointing continuity without adding hardware.

Localyte is being validated across different sensor qualities and use cases, from low-cost IMUs to higher-grade inertial sensors. The long-term goal is a plug-and-play inertial enhancement layer: installable, configurable, and compatible with existing navigation stacks.

My work focuses on AI-based correction models, sensor-fusion ideas, validation pipelines, and the translation of research prototypes into deployable software.

I am interested in systems that survive messy physical conditions: noisy sensors, imperfect models, missing references, and changing environments. Localyte is a way to bring that research perspective into a real product: improving the signals that robots, drones, vehicles, and satellites depend on to understand their own motion.