About Voltspan
Voltspan helps companies keep their data in their own hands and turn it into models of their own. We build from the power up: our own GPU clusters in US data centers, the sandboxes and inference agents run on, and the training behind them. We are growing across every layer of the stack.
About the role
Agent traffic is bursty, long-context and full of tool calls. You will build Token Factory, the inference service that serves open and customer-trained models with steady latency under that load.
What you'll do
- Build and tune serving for open models on GB300 and B300
- Improve KV cache reuse, batching and routing for multi-turn agent sessions
- Serve many small customer models efficiently on shared hardware
- Measure throughput, latency and cost per token, and push all three
What we're looking for
- Experience with LLM inference engines (vLLM, SGLang, TensorRT-LLM or similar)
- Strong Python and C++ or Rust
- Solid understanding of GPU memory and batching trade-offs
Nice to have
- Speculative decoding or prefix caching work
- Multi-LoRA serving
Projects you might work on
- Serve hundreds of fine-tuned models from one cluster without latency spikes
- Prefix caching that cuts the cost of long agent sessions in half
Logistics
San Jose, CA · Hybrid · Full-time
If a role excites you but you don't meet every qualification, apply anyway. We care more about what you have built than about a checklist.
Salary range, visa sponsorship and benefits will be added before this posting goes live.
How we work
- Small teams, big systems. Everyone here owns real infrastructure and real customers from week one.
- From volts to agents. Work moves across the whole stack, so you will see how your piece connects to power, GPUs, tokens and agents.
- We run what we build. The people who write the software also operate it, next to the people who rack the hardware.
Equal opportunity
Voltspan is an equal opportunity employer. We consider all qualified applicants without regard to race, color, religion, sex, sexual orientation, gender identity, national origin, age, disability, veteran status or any other characteristic protected by law. We consider qualified applicants with arrest and conviction records in accordance with the California Fair Chance Act.
If you need an accommodation during the application process, let us know.