Foundational VLM for Robotics Perception

Fine‑tuned CLIP with 3D point‑cloud embeddings and GPT‑style prompts for spatial scene understanding

Real‑Time CUDA Point‑Cloud Filter

Custom CUDA kernels integrated with ROS 2 for low‑latency filtering & publish‑subscribe pipelines

Azure MLDC Automation Suite

End‑to‑end pipeline to version datasets, orchestrate training, & deploy TensorRT models from Azure DevOps

Autonomous PTZ Barcode Scanner

Auto‑orients a 3‑DOF PTZ camera to detect & decode barcodes on moving freight in real time

Pallet Detection & Pose Estimation

Deep CNN locates pallet pockets and estimates 6‑DoF pose to guide autonomous forklifts

LiDAR‑Camera Freight Stacking

Fuses 3D LiDAR with RGB tracking to autonomously stack automotive freight with sub‑inch accuracy

Warehouse‑Scale 3D Reconstruction

Voxel‑hash mapping pipeline reconstructs 50 m³ environments at 15 fps for inventory analytics

Ground‑Plane Segmentation ROS Stack

CUDA‑accelerated ground removal node processes 300 kpts/s, boosting downstream object detection recall by 12 %

Safety‑Alert Zone Monitoring

Eigen‑based proximity algorithm triggers real‑time alerts when humans enter AGV danger zones

2D‑3D ToF Camera Calibration Toolkit

Optimizes extrinsics between ToF depth and RGB, reducing avg reprojection error to 0.8 px