About the job
To achieve this, we are developing a robust hardware-software ecosystem leveraging multi-modal wireless mesh sensing technology. This innovation allows us to significantly reduce the cost and time involved in sensor deployment by a factor of ten. Ultimately, our platform aims to serve as the perception engine for businesses, facilitating real-time visibility and autonomous management of their operational perimeters.
Our co-founders, Xerxes and Philip, are deeply committed to empowering our partners in the rapidly evolving landscape of physical AI and robotics. We are a dynamic, rapidly expanding team comprised of talent from Anduril, Tesla, Uber, and the U. S. Special Forces.
Position Overview
Specter is seeking a dedicated Machine Learning Infrastructure Engineer to construct and optimize the ML systems that drive real-time perception and inference capabilities across our edge-cloud platform. This position will involve overseeing the training, deployment, and enhancement of computer vision and sensor fusion models, aimed at enabling autonomous monitoring and decision-making for our clients' physical assets.
Key Responsibilities Include:
Design and implement scalable ML training pipelines for computer vision applications, including object detection, tracking, classification, and segmentation.
Develop efficient model serving infrastructures to facilitate real-time inference on edge devices with limited computational and power resources.
Optimize models for deployment on embedded hardware, employing techniques such as quantization, pruning, TensorRT, ONNX, and CoreML.
Create continuous training and evaluation systems to enhance model performance through feedback loops derived from production data.
Establish data pipelines for the ingestion, labeling, versioning, and management of extensive multi-modal sensor datasets, including video, radar, lidar, and thermal data.
Implement model monitoring frameworks, A/B testing methodologies, and performance analytics for deployed perception systems.
Collaborate with perception researchers to transition models from research environments to scalable production across thousands of edge nodes.
Construct tools and infrastructure for distributed training, hyperparameter optimization, and experiment tracking.
