Software for AI-Driven Industrial Applications
Narrow AI systems for robotic perception, task interpretation, and reliable execution. Focus on architectures that are testable, integrable, and suitable for industrial constraints.
Areas of focus
Robotic perception
Practical scene understanding for structured and semi-structured environments, built for robustness rather than demos.
Decision logic & execution
Constrained, task-specific logic that bridges perception outputs to predictable robot actions and recovery behavior.
Simulation-driven development
Validation of task logic and failure modes in simulation to reduce iteration cost and deployment risk.
Integration mindset
Design that respects real interfaces (controllers, PLCs, industrial networks) and keeps AI out of safety-critical loops.
Selected projects
Short, technical summaries with emphasis on system behavior and constraints. Implementation details are intentionally omitted.
Jetank: Perception-Guided Toy Picking
Demonstration of perception-driven object selection and pick execution on resource-limited hardware.
Adaptive Palletizing Strategy in Simulation
Simulation-based development of a palletizing strategy that reacts to the evolving pallet state and constraints.
Language-Guided Robot Task Interpretation
Using language models as an interpretation layer to map intent and scene descriptions into a predefined action vocabulary.
Sim-to-Real Fruit Detection with Synthetic Data
Evaluation of synthetic, real, and hybrid datasets for fruit detection, including deployment on embedded hardware.
Robot Task Orchestration with Behavior Trees
Modular orchestration of robotic workflows with explicit sequencing, fallback behavior, and separation between decision logic and action layers.
Sim-to-Real Execution with Torobo
Controlled pick-and-place execution using inverse kinematics in NVIDIA Isaac Sim as a baseline for perception-driven manipulation.
