Etched is building the first transformer-specific AI chip. By trading flexibility for performance, one Etched server has >20x more throughput (tokens per second) than an NVIDIA 8xH100 GPU. We've raised >$125M and are a 60-person team in Cupertino, CA.
From spending 30% of your time doing research to discussing recent ML papers with customers to demo'ing performance models for how our chip will run their workloads, you will be the technical expert working with customers as we ramp to $100M in pre-orders.
Representative Projects
Presenting to the CTO, Head of AI, and Hardware/Data Center expert at a AI company, cloud, or trading firm to understand their needs and demonstrate the value of Sohu
Running and improving customer demos of our product
Giving technical feedback to the engineering and ML teams to improve the product
Creating Sohu cluster configurations given customer requests
AI research
Ideal Traits
Great with people: builds trust, communicates effectively, and understands client dynamics
Technical depth: deeply understands complex technical concepts, gives a "wow factor" when presenting to engineering teams
Credibility: understands the customer problems (ideally experienced them firsthand)
Ideal Experience
Deeply familiar with transformer models and inference optimization
At least somewhat familiar with AI hardware and data center infrastructure
Customer-facing and/or partnerships experience
Bonus Points
Worked at a potential Etched customer (AI company, cloud, or trading firm)
Wrote custom kernels, evaluated different AI chips, and/or led compute decisions for an AI company