Summary
✨ AI‑Generated
Become a founding engineer focused on making AI inference infrastructure faster and more efficient. You will monitor bare-metal GPU systems, tune modern model-serving stacks, measure throughput and energy efficiency, and build intelligent scheduling and routing software across large-scale compute environments.
Highlights
Work at the intersection of AI inference, GPU infrastructure, networking, and performance optimization, with direct ownership of production systems and opportunities to work closely with data center operations teams.
Description
Emotos builds the control layer for agentic AI factories.
Our product decides which model handles each agent task, schedules agent teams across GPUs, and moves state across the cluster so the same hardware gets more work done.
Think of it as Airbnb for data centers: tasks are the guests, GPU capacity is the stay.
We already run a production model gateway, a request classifier and a benchmark harness that measures GPU energy per successful task.
What you'll do
Monitoring bare-metal GPU nodes: drivers, CUDA, NCCL, InfiniBand/RoCE, NVLink, storage.Deploy and tune the serving stack (vLLM, SGLang, NVIDIA Dynamo, llm-d), then run our software on top: router, classifier and agent-aware scheduler.Measure what matters: GPU, node and rack power; KV-cache use; throughput; successful tasks per GPU-hour.
Compare against the operator's baseline, with repeats, and write it up so their engineers trust it.Work directly with the data center's ops and network teams, then take the lessons back into the product.Extend scheduling across racks and optical fabrics (co-packaged optics, optical circuit switching) so agent state can move where it's cheapest to run.
You have
Served LLMs in production at scale: batching, KV cache, prefill/decode, multi-GPU and multi-node parallelism.Run your own GPU clusters (Linux, Kubernetes or Slurm), not only managed endpoints.Rigour: you know the difference between a benchmark and a proof.Strong Python plus one systems language (C++, Rust or Go).
Ideal Candidate
Data-center networking or optics experience: 400G–1.6T interconnects, RDMA, co-packaged optics, optical circuit switching.KV transfer and offload (NIXL, LMCache, Mooncake), or power and energy measurement (DCGM, PDU/IPMI).
$170k–250k base + 0.75–7.5% equity, founding role.
If you fit the description, we would love to talk with you.
