Embedded CV Engineer — Edge ML & Real-Time Inference

Mecka; New York (NY); On-site; USD 120,000 - 134,000; Full time

Job summary

Mecka is seeking engineers to deploy advanced computer vision on embedded hardware, delivering efficient, low-latency solutions for robot learning. You will train models, run them on constrained devices, and push accuracy from controlled tests into real-world operation.

The role includes building C++ inference pipelines, optimizing for latency, memory, and power, and collaborating across teams to ensure robust deployment and evaluation.

Qualifications

• Strong Python and C++ programming skills with debugging and performance profiling.
• Experience training and evaluating computer-vision models with PyTorch.
• Experience developing on embedded Linux and deploying models on resource-constrained hardware.
• Understanding of quantization, model conversion, and inference optimization on CPUs, GPUs, or NPUs.
• Ability to evaluate accuracy, false positives, and missed detections across changing conditions.
• For the lead position: experience making architecture decisions and guiding engineers.

Responsibilities

• Train and evaluate computer-vision models for deployment on embedded devices.
• Build efficient C++ inference pipelines and integrate with embedded Linux software.
• Optimize models and runtime performance for latency, memory, power consumption, and sustained operation.
• Test performance on representative data and hardware, and investigate errors across preprocessing, inference, and integration.
• For the lead position: set architecture and evaluation standards while continuing to develop models and software.

Skills

Python

C++

Performance profiling

Model training/evaluation

Embedded Linux

Inference optimization

Quantization

Architecture decisions

Leadership

Tools

PyTorch

ONNX

Job description

Mecka is seeking engineers to deploy advanced computer vision on embedded hardware, delivering efficient, low-latency solutions for robot learning. You will train models, run them on constrained devices, and push accuracy from controlled tests into real-world operation.

The role includes building C++ inference pipelines, optimizing for latency, memory, and power, and collaborating across teams to ensure robust deployment and evaluation.