Field CTO, Inference & Networking

Field CTO, AI Inference & Networking Scala Computing · Senior team · Reports to the CEO · Remote · Full-time Why Scala Scala Computing helps companies measure how AI workloads perform on networks at scale. Model labs, neoclouds, silicon vendors, network equipment makers, and the teams that run large training and inference clusters use Scala to answer expensive questions before they commit to hardware, and to understand why clusters that are already running deliver less than their hardware should. Workloads enter the simulator as measured traces, and results leave it as evidence a customer can act on. The person in this seat gets an unusual view of the industry. You will see how some of the largest AI clusters are designed before they are built, and you will help shape the decisions that determine how they perform. The role The Field CTO is Scala’s senior technical leader in front of customers and the industry. You will work with customers’ ML engineers and network architects as a peer, decide with them what a simulation engagement should answer, lead the software that turns simulation output into findings they can act on, and serve as Scala’s public technical voice. You will report to the CEO and serve on the senior team. Implementation engineers sit in the core engineering organization. Depending on how the team evolves, you may or may not carry direct reports, but you will set technical direction either way. What this role is not This is not a pre-sales or sales engineering role, a professional services role, or a marketing or evangelist role. You will own the technical substance of what Scala tells its customers and what it builds for them. What you will do • Executive technical partnership. You will lead technical relationships with customers’ infrastructure, ML platform, and networking leaders. You will run scoping sessions, frame the questions a simulation study needs to answer, and present findings to engineers and executives. • Rigorous requirements discovery. Customers often arrive with a symptom rather than a question, such as a cluster that delivers less than expected or a fabric decision they cannot test on a bench. You will turn that symptom into a defined experiment with workloads, topologies, and success criteria. • Owning the workload side of simulation. You will bring a deep understanding of how LLMs are trained and how they serve tokens, including data, tensor, pipeline, and expert parallelism, the collective communication those strategies produce, prefill and decode, KV-cache behavior, and batching and scheduling. You will own how those workloads are represented in the simulator, so that what we model reflects what production systems actually do. • Owning the networking substance. You will bring depth on scale-out fabrics, scale-up interconnects, and how the two interact. You will be fluent in the mechanisms that decide how a fabric behaves under collective traffic, such as adaptive routing, packet spraying and trimming, congestion control, priority flow control and ECN, multipathing, and in-network telemetry. • Building the results loop. You will lead the design of software that automates the path from raw simulation output to customer-facing findings, so that results reach customers as a repeatable product rather than a bespoke report. • Guarding credibility. You will establish validation practice built on hardware measurements, telemetry, vendor specifications, and customer-provided data. You will decide what the evidence supports, and you will keep the line between demonstrated capability and roadmap honest in customer conversations and in public. • Internal influence. You will bring what you learn from customers back to engineering and product, make scoping and feasibility calls on engagements, and help decide what Scala commits to and when. • Thought leadership. You will publish white papers, speak at conferences, appear on podcasts, and represent Scala in ecosystem and standards efforts such as MLCommons and the Ultra Ethernet Consortium. What we expect • Experience at scale. You have 10 or more years in AI infrastructure, data center networking, or distributed systems, including time at a hyperscaler, a frontier AI lab, a neocloud, or a major silicon or network equipment vendor. Time inside a model lab or a neocloud is especially valuable, because those are the customers this role serves. • Executive presence. You can hold your own with network architects and ML engineers in the same meeting, and you can draw out requirements from people who have not yet fully articulated them to themselves. • A forensic mindset. You can tell a theoretical model apart from the way real GPUs, NICs, switches, and production workloads behave, and you insist on the difference. • Software leadership. You are hands-on enough to prototype and to review implementation work credibly, and you can direct engineers without being their manager. • A public track record. You write and speak clearly in public. Published papers, standards contributions, or conference talks are strong evidence. AI infrastructure expertise • You understand LLM training and inference at the level of the algorithms. You know how a training step becomes gradient and activation traffic, how inference produces tokens, and how architectural choices such as mixture-of-experts change the traffic a cluster carries. • You know data center networking down to NIC and switch internals, including queues, buffers, and schedulers, as well as RDMA and RoCE, congestion control, and the contention and failure behavior of large fabrics. • You have direct, hands-on experience with the NVIDIA platform, including Spectrum-X, NVLink and NVSwitch, and InfiniBand, and you know how NVIDIA GPUs and their collective libraries behave on each of those fabrics. This is a firm requirement for the role. Nice to have • You have worked with discrete-event network or system simulation, or with performance modeling of large systems. • You know the workload-trace ecosystem, including Chakra execution traces, PyTorch profiling and execution traces, and MLPerf benchmarks. • You have participated in MLCommons, the Ultra Ethernet Consortium, OCP, or IETF work, or contributed to open-source networking or simulation projects. Compensation $245 -325K + Equity based on experience. Location and travel This is a remote role, Bay Area preferred. You will travel for customer work and conferences, roughly 10-15% of the time. How to apply • Send a note about the largest system you have helped run, design, or explain to careers@scalacomputing.com. If you have written papers or given talks, include links to them.