Research Engineer, Inference

Plsn, Palo Alto, California

What you will do

Own the path from a research checkpoint to a deployed model. Build the serving infrastructure and make new models straightforward to ship.
Profile and improve latency, throughput and cost across the inference stack. Measure the bottlenecks, make a change, and prove that it helped.
Evaluate quantization, distillation, KV-cache optimization and batching where they fit the model. Keep correctness and model quality in view as you optimize.
Build and tune serving systems for low-latency, real-time and streaming workloads. Own scheduling, memory use and GPU utilization.
Write or tune GPU kernels where the existing stack falls short. Work from model architecture down to the kernel and back up to orchestration.
Build tools that let researchers deploy models quickly and safely. Make performance regressions visible before a model ships.

Plsn, Palo Alto, California

$150K–$220K

a year

This base pay is below the median for AI research jobs in the Bay Area with no level in the job title.

We compare the middle of each posted range. This is base pay as posted. The group does not include this job.

The Plsn site opens in a new tab.

Compare your résumé with this job. See where you match, what is missing and which lines to change.

Own the path from a research checkpoint to a deployed model. Build the serving infrastructure and make new models straightforward to ship.
Profile and improve latency, throughput and cost across the inference stack. Measure the bottlenecks, make a change, and prove that it helped.
Evaluate quantization, distillation, KV-cache optimization and batching where they fit the model. Keep correctness and model quality in view as you optimize.
Build and tune serving systems for low-latency, real-time and streaming workloads. Own scheduling, memory use and GPU utilization.
Write or tune GPU kernels where the existing stack falls short. Work from model architecture down to the kernel and back up to orchestration.
Build tools that let researchers deploy models quickly and safely. Make performance regressions visible before a model ships.