Multi-sensor fusion
LiDAR, mmWave radar, ultrasonic and camera streams are hardware-timestamped, extrinsically calibrated to a common vehicle frame and aligned to a single time base, then fused into a shared spatial representation.
Multi-sensor fusion and on-vehicle AI model deployment.
OmnixDrive performs intrinsic and extrinsic calibration and temporal alignment across LiDAR, mmWave radar, ultrasonic and camera streams, fuses them into a shared spatial representation, and trains hardware-aware models for on-vehicle inference on automotive SoCs.
- Founded
- 2026
- Headquarters
- Singapore
- Focus
- Sensor fusion · on-device AI
Fusion, on-device models and the calibration loop
The platform covers sensor calibration, temporal alignment, multimodal fusion, hardware-aware model training and deployment on automotive SoCs.
Dedicated on-device models
Perception and planning models are trained against the fused representation, then quantised, distilled and pruned to meet a fixed latency, memory and power budget on the target automotive SoC.
Calibration & data loop
Fleet data returns to the pipeline: auto-labelling, calibration drift monitoring and versioned datasets keep the model bound to the sensor configuration it was trained on.
Standards & compliance
Our development process is defined against the standards regulated automotive programmes require — functional safety, software process, information security and data protection.
- ISO 26262 Functional safety process, ASIL-D capable
- ISO 21448 Safety of the intended functionality
- ASPICE L2 Automotive software process assessment
- ISO 27001 Information security management
- TISAX Automotive information security assessment
- GDPR EU data protection requirements
Perception runs on the vehicle; training runs in your data centre, a private cloud or ours.
Production operating targets
- 4 Sensor modalities fused on one time base — LiDAR, radar, ultrasonic, camera
- ≤ 50 ms On-device inference budget from fused perception to planning
- PB+ Multimodal driving data under management per programme
- 99.95 % Platform availability target
Company, platform and contact
Technical platform, programme applications, company information, careers and contact details.
About
Company profile, engineering scope, operating principles, team and locations.
Read about OmnixDrive 02Technology
The fusion pipeline, the on-device runtime and the deployment models we support.
See the platform 03Solutions
OEM and Tier-1 programmes, sensor-supplier evaluation, and the engagement models available.
Review solutions 04Careers
Engineering disciplines, working model, employment terms and application details.
View open roles 05Contact
Business, technical, partnership, press and recruitment contact details.
Contact us
Perception pipelines and on-vehicle models
Founded in Singapore in 2026, OmnixDrive develops multi-sensor perception pipelines and dedicated on-vehicle AI models for autonomous driving. Its engineering scope covers multimodal perception, sensor calibration and geometry, model compression, embedded inference and dataset operations.
Calibration and temporal alignment
Intrinsic and extrinsic calibration, coordinate transformation and time-base alignment are validated before model training begins.
On-vehicle runtime
Model training and optimisation are performed against the latency, memory and power budgets of the target automotive SoC.
Evaluation and traceability
Accuracy is reported per sensor and per scenario, with model, dataset and calibration versions retained for regression analysis.
Technical consultation
A technical consultation covers the sensor configuration, calibration chain, target automotive compute and scope of an initial deployment phase.