Senior AI Research Engineer

📋 Description

• • Plumerai is at the forefront of enabling developers to integrate highly accurate Artificial Intelligence into embedded devices, empowering them to create groundbreaking new products. Our innovative approach combines on-device Tiny AI software with sophisticated cloud-based multimodal Large Language Models (LLMs). This powerful synergy delivers advanced features such as People Detection, Video Search, Familiar Face Identification, and AI Captions, transforming the capabilities of smart devices.
• • Major enterprises are already deploying our cutting-edge computer vision models across millions of smart home cameras globally. We are experiencing rapid expansion into critical sectors including commercial security, retail, and assisted living, demonstrating the broad applicability and impact of our technology.
• • Our core philosophy centers on maximizing on-device processing to ensure AI solutions are low-power, highly accurate, and inherently private. Plumerai distinguishes itself by achieving superior accuracy, even surpassing established players like Google Nest in certain benchmarks. We are now strategically investing heavily in market adoption to accelerate our growth and reach.
• • We achieve state-of-the-art accuracy and efficiency in AI solutions through a vertically integrated approach, meticulously controlling every layer of the technology stack. This includes data collection and curation, custom training software development, innovative model architecture design, the creation and refinement of multimodal LLMs, sophisticated pre- and post-processing techniques, and the optimization of high-performance inference engines.
• • Our team possesses a profound theoretical understanding across all these critical components, coupled with a proven ability to ship products rapidly and frequently. This blend of deep expertise and agile execution is fundamental to our success.
• • As a Senior AI Research Engineer, you will play a pivotal role in developing and advancing our state-of-the-art AI products. Your responsibilities will span a wide spectrum, from enhancing our sophisticated training algorithms and integrating advanced multimodal LLMs to building robust data pipelines, architecting novel model designs, and applying well-established machine learning methodologies with creative problem-solving.
• • You will be instrumental in building new AI features that are destined for deployment on millions of camera devices in the field, directly contributing to the development of the most advanced AI for embedded systems.
• • Specifically, you will leverage and enhance multimodal LLMs to unlock new functionalities for our customers and optimize their deployment across both cloud and edge environments. This involves pushing the boundaries of what's possible with LLMs in resource-constrained settings.
• • You will tackle the challenge of designing and training increasingly accurate deep learning models, particularly focusing on those with exceptionally small memory footprints – our smallest computer vision model is a mere 1MB. This work enables new, complex AI applications to run efficiently on low-cost, low-power hardware, democratizing advanced AI capabilities.
• • Your contributions will extend to improving our data pipeline, refining model architectures, and enhancing our training software. While we leverage existing literature, we frequently require novel approaches and ingenious solutions to overcome the unique challenges inherent in our work.
• • You will utilize our robust infrastructure, including a Kubernetes cluster for PyTorch and TensorFlow training jobs, Snowflake and Dataflow for dataset construction, and tools like Streamlit for rapid prototyping of new demos. Extensive GPU resources on Google Cloud Platform (GCP) will be employed for training new models and automating data labeling processes.
• • This role offers a unique opportunity to work on impactful AI research that directly translates into real-world products used by millions, contributing to the future of intelligent embedded devices.