Physical AI & Robotics

522ns from query to PyTorch tensor. Built for real-time ML pipelines.

---

Why Robotics Teams Choose ZeptoDB

522ns to Tensor

Zero-copy path from column store to NumPy/PyTorch. No serialization bottleneck.

ASOF JOIN

Time-align heterogeneous sensor streams (LiDAR, IMU, camera) with nanosecond precision.

5.52M events/sec

Ingest 10KHz+ sensor data from multiple sources without backpressure.

Parquet HDB

Replay historical episodes from S3 for offline training and evaluation.

---

Use Cases

Sensor Fusion

• ASOF JOIN to time-align LiDAR point clouds, camera frames, and IMU readings
• Nanosecond-precision timestamps for accurate temporal correlation
• Window JOIN for ±N ms tolerance matching

RL Feature Store

• Real-time feature computation: rolling stats, EMA, percentiles
• Zero-copy to PyTorch tensors for online inference
• Historical replay from Parquet HDB for offline training

Fleet Telemetry

• Ingest telemetry from thousands of robots/vehicles simultaneously
• Real-time anomaly detection with window functions
• Multi-node cluster for fleet-scale data

---

Example: Sensor Fusion Pipeline

-- Align LiDAR scans with nearest IMU reading and camera frame
SELECT
    l.ts,
    l.point_cloud,
    i.accel_x, i.accel_y, i.accel_z,
    i.gyro_x, i.gyro_y, i.gyro_z,
    c.frame_id
FROM lidar_scans l
ASOF JOIN imu_readings i ON l.robot_id = i.robot_id AND l.ts >= i.ts
ASOF JOIN camera_frames c ON l.robot_id = c.robot_id AND l.ts >= c.ts
WHERE l.ts > now() - interval '10 seconds'

import zeptodb
import torch

conn = zeptodb.connect("localhost:8123")

# Query → NumPy → Tensor in microseconds
result = conn.query("SELECT accel_x, accel_y, accel_z FROM imu_readings WHERE ts > now()-1s")
tensor = torch.from_numpy(result.to_numpy())  # zero-copy

---

Integration Points

System	Integration
ROS2	Plugin (roadmap) — direct topic subscription
MQTT	Ingestion connector for IoT sensors
OPC-UA	Industrial PLC connector
PyTorch	Zero-copy via NumPy bridge
Arrow Flight	Batch streaming for training pipelines