Cyber Manufacture Co. is an AI company developing a personalized operating system that feels more intuitive, friendly, and powerful. Our key technologies, kernel™ and rabbit™, speed up multimedia language model applications by an order of magnitude and enable real-time, application-programming-interface-free agency on any platform.

We are a small, fast-moving team based in Los Angeles. We are well-funded and backed by Khosla Ventures, led by Vinod Khosla, co-founder of Sun Microsystems.

At Cyber Manufacture Co, we believe that the best way to realize the value of cutting-edge research is by creating delightful, end-user-oriented experiences with taste. We pride ourselves on being uniquely positioned to capture the opportunities during this technology renaissance with our product focus. Our team has created mature software, hardware, and designs that have impacted millions of people, making us one of the few capable of all components of the vertically integrated chain.

We value individuals with:

• a deep curiosity about technology
• out-of-the-box problem-solving skills
• an understanding of building and operating complex systems in production
• an obsession with customers and crafting experiences
• a sense of responsibility and teamwork

The machine learning team at Cyber Manufacture Co. prototypes and builds core machine learning systems that make our operating system experience unique, ranging from large language models to multi-modal foundation models. Your work will be both pragmatic and pioneering: the models you develop will support the day-to-day functioning of our operating system in areas such as computer vision, conversational intelligence, speech recognition, and synthesis. Furthermore, your work will expand our understanding of how AI interacts with the world in a way that is similar to human interaction.

• Maintain, iterate, and deploy machine learning algorithms in production on our cloud systems.
• Research and propose novel methods for some of our most challenging problems that require a machine-learning-based solution. Oversee experimental progress through large-scale training and benchmarking pipelines, and formalize them into alpha and beta features of our operating system.
• Evaluate third-party machine learning services and cloud computing vendors for their performance. Provide a perspective on which to build in-house, which to benefit from external partnerships.
• Independently execute tasks, solve problems, and report to technical leadership, with the goal of leading a broader team of full-time employees and interns in the future.

Ideal candidates should be familiar with at least two parts of the full lifecycle of model development:

• Acquiring large-scale data, either through scraping or proprietary acquisition, of traditional multi-modal data (text, image, code) and data of a more esoteric nature.
• Curating subsets of large datasets to meet product needs while remaining IP compliant, and serving such data to the training compute infrastructure.
• Understanding and characterizing the behavior and limitations of existing models.
• Developing, implementing, and testing plans to improve such models with cutting-edge, original research.
• Utilizing distributed training frameworks on large compute clusters with multi-node GPU and high-speed interconnect.
• Tuning inter-node bandwidth over communication protocols like RDMA over Converged Ethernet (RoCE) with software integrations including NVIDIA Collective Communications Library (NCCL), Gloo, or Message Passing Interface (MPI).
• Developing training and fine-tuning pipelines for large models at scale.
• Optimizing and deploying models on a variety of homogeneous and heterogeneous hardware on the cloud or at the edge.

This role is ideal for candidates who have two or more years of industry experience.

Here is a brief overview of the models we are using, training, and deploying. Any experience working with them is valued and appreciated:

• Large language models for conversation, structured data processing, tool usage, and reasoning
• Streaming speech recognition models
• Streaming speech synthesis models (concatenative and neural)
• Empowering language models with long-term memory and real-time information: Retrieval Augmented Generation, integration with vector database
• Multi-modal foundation model for interface understanding (UI & UX), from rasterized image to DOM
• Named entity recognition, utterance and sentence segmentation, and zero-shot text classification and embedding models
• Open-world instance segmentation and object-tracking models
• Landmark detection of man-made and natural structures (tourist attractions, vendors)
• Object re-identification models

This role is based in our Los Angeles headquarters. However, we may consider remote candidates in exceptional circumstances.

We want you to be an owner in our company and share in executing our vision, so all full-time employees receive equity. In addition, we offer competitive market compensation, medical and dental/vision benefits, flexible vacation, and an opportunity to work for a uniquely positioned, VC-backed company with significant upside potential. Cyber Manufacture is committed to fair and equitable compensation practices. The annual salary range for this role is 150,000 to 200,000. The final package may vary depending on job-related knowledge, skills, candidate location, and prior work experience; the total compensation package for this role also includes stock options.

At Cyber Manufacture, we are committed to creating an inclusive and diverse workplace. We welcome and encourage applicants from all backgrounds, and all qualified applicants will receive consideration for employment without regard to race, color, religion, gender, gender identity or expression, sexual orientation, national origin, genetics, disability, age, veteran status, or any other protected class.