Image Algorithm Engineer

About Company

Founded in 2024, Acrab is on a mission to build reliable, accessible AI agents for everyday life and creative work. By developing a full-stack architecture spanning AI silicon, local LLM inference, operating systems, multi-modal human-machine interfaces, and agent orchestration, Acrab brings AI from the cloud to the edge, enabling intelligent agents to be seamlessly embedded into the world around us.

Role Summary

As an Image Algorithm Engineer, you will develop and optimize computer vision algorithms for embedded platforms, while exploring VLM/LLM and Agent applications to enable intelligent multimodal interaction and real-time on-device AI.

Key Responsibilities

• Develop and optimize computer vision algorithms, including DMS, AVM, Sentry, and privacy/de-identification features.
• Design and optimize lightweight vision algorithms for embedded platforms, addressing low compute, low latency, and high-reliability requirements.
• Build and manage the end-to-end vision data pipeline, including data collection, annotation standards, data cleaning, augmentation, model training, deployment, and iteration.
• Explore integration applications, combining perception, external environment sensing, and decision-making.
• Develop VLM/LLM-powered applications, including visual intent understanding, multimodal assistants, and real-time environmental information extraction.
• Explore vision-based AI Agents for proactive decision-making and context-aware interaction based on human and environmental states.
• Deploy and optimize VLM/LLM models on edge chips, including model compression, inference acceleration, and heterogeneous computing for low-latency, low-power inference.

How You Will Stand Out

• 3+ years of experience in computer vision algorithm development, with at least 1 year of experience such as DMS/OMS/IMS is preferred.
• Strong knowledge of object detection, keypoint detection, segmentation, tracking, and deep learning frameworks such as PyTorch or TensorFlow.
• Hands-on experience with model training, optimization, and deployment on embedded platforms.
• Experience with ONNX, TensorRT, TFLite, or NPU toolchains for model conversion and acceleration.
• Experience integrating perception with decision-making or autonomous systems.
• Experience with VLM/LLM applications, multimodal models, fine-tuning, or on-device inference optimization.
• Experience developing AI Agents, including task planning, memory, tool calling, streaming, or multimodal perception and decision-making.