Lead Research Engineer, Applied AI

About us

At Ethosphere we are combining hardware, software and AI technologies to build an AI-powered coaching engine for frontline retail employees. We are partnering with major retail brands to engage, develop and grow retail associates into exceptional sellers, who feel seen, heard and celebrated in the flow of their work. Ethosphere believes in a world where AI gives humans superpowers to achieve greatness, while deepening our connections to one another, as humans.

Our Working Ethos

Our mission is to help companies elevate and celebrate the people who bring their brands to life every day. As a team, we come together to innovate, ideate and iterate to collectively capture the potential that AI holds for every individual. We invite dissent in the pursuit of excellence and approach every business opportunity with a growth mindset, while celebrating our collective and individual successes.

About our Shop

We’re tackling ambitious challenges on an exciting roadmap, developing software that operates at the edge and in the cloud. Our backend services run on Kubernetes, utilizing cloud-based APIs from AWS and GCP. We focus on building core, strategically important technology while leveraging cloud services and open-source solutions where it makes sense. To support our product, we work with signal processing algorithms, advanced speech separation and recognition models, large language models (LLMs), retrieval-augmented generation (RAG), and agent-based systems. Primarily a Python shop, we also dive into C, C++, Go, and CUDA to tackle complex problems as needed.
We’re looking for talented, driven engineers to join our team early on. We value passion, commitment to high-quality work, constant learning, and humility to collaborate and learn from others. Here, you’ll have significant ownership and direct impact on all aspects of our product. We’ll support, mentor, and help you grow in your career along the way.

About the Role

As the Lead AI Research Engineer at Ethosphere, you will drive the development and application of advanced AI technologies across all aspects of our engineering stack that will help provide impactful and personalized coaching for frontline employees at scale.

In this role, you will:

Design and implement advanced neural network models for audio separation, enhancement, and related speech processing tasks, leveraging transformer-based models, and convolutional neural networks (CNNs).
Lead the effort to optimize the performance of these models on various hardware platforms including Nvidia / AMD GPUs and Qualcomm NPUs.
Design, implement, and refine intelligent agents within an AI-powered coaching engine for frontline retail employees, using a complex, multi-step agentic LLM analysis framework that leverages contextual learning, memory, and retention to deliver personalized and adaptive coaching and support ongoing employee growth and engagement.
Implement a lifecycle approach to model development by continually enhancing model robustness and performance, using iterative improvements that adjust based on real-world interactions and performance metrics.
Maintain and optimize codebases in Python and PyTorch to enable scalable and cost effective inference strategies.

You may be a good fit if you have:

A Master's/ PhD degree in Computer Science, Machine Learning, Mathematics, Statistics or a related field.
A strong foundation in deep learning, with a focus on transformer models, convolutional neural networks, self-attention mechanisms, audio signal processing, or natural language processing.
Proven experience in designing, training, and optimizing neural network models, particularly for audio and speech separation tasks, using frameworks like PyTorch.
Hands-on experience in training, fine-tuning and deploying large language models.
Proficiency in Python and experience with machine learning libraries and toolkits such as NumPy, SciPy, and Librosa, with familiarity in integrating CNNs with other architectures like Hugging Face Transformers.
Hands-on experience with audio data preprocessing, spectrogram generation, and augmentation techniques to improve model robustness.
Advanced knowledge of computational efficiency, deploying low-latency models for edge applications on Qualcomm GPUs or NPUs.
A background in research and development for machine learning, ideally demonstrated through publications, projects, or contributions to open-source repositories.
Familiarity with recent advancements in agentic LLM development and a passion for applying agentic learning and memory mechanisms to improve AI-driven interactions and engagement.
Have prior experience working with vector databases, search indices, or other data stores for search and retrieval use cases.
Familiarity with GitHub and collaborative code development practices, including maintaining clear documentation, writing modular code, and implementing reproducible experiments.
The ability to work effectively in a fast in an environment where things are sometimes loosely define