SlipstreamJobs tracks this role from the company's public career site. Apply directly on the employer's site.
You will own the cost and performance of Adaption's inference stack, determining how efficiently the company serves models as workloads, traffic, and hardware change. Working closely with engineers operating the serving fleet, you will control the core performance levers: caching, batching, quantization, decoding, and kernel-level optimization. Success is measured by improving throughput and latency without compromising reliability or model quality.
Key responsibilities include:
- Improve throughput, cost, and tail latency through KV-cache management, continuous batching, speculative decoding, and quantization
- Optimize long-context prefill and decode workloads based on real production traffic patterns
- Tune routing between internal infrastructure and external providers based on cost, capacity, and performance trade-offs
- Work within serving engines such as vLLM, SGLang, and TensorRT-LLM, diving below the framework level when needed for optimization
- Build profiling and measurement systems that expose where time, memory, and compute are being spent
Adaption is building efficient intelligence that evolves in real-time. The company's vision centers on AI systems that are flexible, personalized, and accessible to everyone, with efficiency as the key enabler for expanding access and ensuring innovation benefits broadly.
The role offers flexible work arrangements with in-person collaboration in the Bay Area, a distributed global-first team structure, and team offsites. Benefits include an annual travel stipend (Adaption Passport), weekly meal allowance, comprehensive medical benefits, and generous paid time off.
REQUIREMENTS:
- 5+ years in ML systems, inference infrastructure, or performance engineering, with measurable improvements in cost or latency
- Deep understanding of model serving, including prefill and decode, memory bandwidth, batching, and concurrency
- Production experience with serving engines such as vLLM, SGLang, or TensorRT-LLM
- Strong Python skills and proficiency in C++, Rust, or another systems language
- Experience with GPU performance, including CUDA, NCCL, mixed precision, memory layout, kernels, or quantization
