Accelerative Synthetic Data Generation
Diffusion-based filtering to detect inauthentic synthetic videos 9× faster with early-exit pipelines. 75% compute savings, 6× faster generation.
Research artifacts, deployed systems, and weird experiments.
A catalogue of projects across foundation models, multi-agent autonomy, and field robotics. Sorted by domain, not by chronology.
02 Generative AI
Diffusion models, synthetic data, and foundation models.
Ortho-Fuse: Orthomosaic Generation
Intermediate optical flow estimation for orthomosaic creation from sparse aerial images. High-quality crop health analysis with reduced overlap (50% vs. 70-80%).
03 Generative AI × Robotics
Combining foundation models with embodied control.
World Foundation Models + VLA
Augmented real-world video datasets for VLA training, addressing data scarcity with minimal trajectories. Enables sample-efficient policy training.
Wildlife Digital Twin
Photoreal NVIDIA Omniverse simulation with generative animal behaviors, herd dynamics, and drone responses. Reduces field deployment costs.
SmartWilds: Multimodal Dataset
Synthetic multimodal dataset for wildlife behaviors using world models. Boosts robust classification from scarce data.
06 Robotics
Autonomous systems, multi-agent coordination, robot control.
Autonomous Vineyard Rover
Segmentation neural network and depth sensing for lane centering on uneven terrain. Robust navigation with 40° slip compensation in real fields.
Multi-Agent RL for Crop Scouting
CNN + RL for heterogeneous agents in agriculture. 60% scouting reduction, 80% accuracy, 4.8× labor savings, 36% farmer profit boost.
Decentralized AV Intersection
Level 3.5 motion planning using road-marking intent detection and graph-theoretic coordination. Communication-free across 255 scenarios.
Priority Patrol Planning
Distributed online patrol with scalable trajectories balancing priority and non-priority site coverage. Sim-to-real validated.
Swarm Synergy Formations
Communication-free community formations for multi-robot swarms. Validated in simulation, experiments, and lab tests.
Heterogeneous UAV Swarm
Multi-agent RL coordination for UAVs in detailed crop health assessment. Scalable deployment without synthetic data dependency.