Skip to main content

Core Technology

A Partner Committed to Outcomes, from the Earliest Technical Decisions

From specification definition, architecture selection, and feasibility studies through to mass-production development, we draw on our development experience and expertise to proactively propose solutions to your challenges. We help reduce unexpected costs caused by rework and shorten development timelines.

Three Technical Gaps Created by the Software-Defined Vehicle (SDV) Era

  • The AI-to-Product Gap

    As AI advances rapidly, the gap between R&D and mass-production development is widening. Optimizing for automotive-specific constraints such as cost, power consumption, and real-time performance is becoming increasingly difficult.

  • The Skills Gap

    Automotive SoCs with hundreds of TOPS are becoming increasingly common, making advanced software expertise essential. Yet across the industry, the skills needed to fully harness them have not kept pace.

  • The Process Gap

    As over-the-air (OTA) software updates become the norm, development must shift from waterfall to agile, and from deterministic, rule-based approaches to statistical, data-driven ones

Solutions

AI Research and Development

  • Technology scouting and assessment
  • AI model design and accuracy improvement
  • Algorithm development
  • Training dataset creation
  • AI development infrastructure

Advanced Development Deployment

  • SoC selection
  • AI model deployment on SoCs
  • Pipeline design and optimization
  • Offloading to hardware IP blocks
  • KPI definition and evaluation

Mass Production

  • System-level performance evaluation
  • Functional safety support
  • Real-time performance verification
  • CI/CD integration

TECHNOLOGY

AI Perception Model Development

We combine the ability to develop perception models in-house from scratch, for tasks such as object detection, lane detection, and traffic sign recognition, with our own data development infrastructure. We can also compress large models, such as Vision-Language-Action (VLA) models. Drawing on implementation know-how specialized for compute-constrained edge devices, we cover everything from the research stage to building production-quality systems.

AI Model Deployment & Optimization on SoCs

Our team has extensive experience deploying and optimizing AI models on SoCs from NVIDIA, Qualcomm, and TI. Through careful task allocation across GPU, DLA, DSP, and VIC, combined with pipeline design, we deliver production implementations that meet automotive-specific constraints.

Software Platform & Quality Infrastructure

We build the development foundations that support production quality, from OS bring-up and functional safety support to automated test frameworks. System verification triggered by code changes through CI/CD integration delivers both development efficiency and quality.

TECH STACK

SoC
NVIDIA DRIVE AGX Xavier / Orin / Thor
Qualcomm QRB5165 / QCS6490
TI TDA4VM / AM62A / AM67A
Hailo Hailo-15H
Telechips N-Dolphin
Realtek
Broadcom BCM2837B0 / BCM2710A1
NVIDIA DRIVE AGX Xavier /
Orin / Thor
Qualcomm QRB5165 /
QCS6490
TI TDA4VM / AM62A / AM67A
Hailo Hailo-15H
Telechips N-Dolphin
Realtek
Broadcom BCM2837B0 /
BCM2710A1
OS
QNX(QOS) / Ubuntu / Yocto Linux / Android
AI Frameworks
PyTorch / ONNX / TensorRT / SNPE / TIDL
Middleware
ROS / ROS2
Programming Languages
C / C++ / Python / CUDA / JavaScript (React/TypeScript)
AI Technology
Object Detection / Segmentation / Depth Estimation / VLM / RAG / Scene Understanding / VLA
Infrastructure
Git / Jenkins / Docker / AWS