PixaSocial Ai builds viral social content on proprietary psychology and psychographic models, published across Instagram, X, LinkedIn and 12+ channels.
THE ROLE
Build the platform that makes our generation stack fast and cheap: GPU utilisation, batching, latency reduction, cost per asset - the infrastructure behind image, video and voice generation.
WHAT YOU WILL DO
- Own inference infrastructure: GPU scheduling and optimisation, batching, caching, model routing, autoscaling.
- Drive down cost per generated asset and p95 latency, with numbers to show for it.
- Build the deployment and observability path for new models: rollout, canary, rollback, health and drift signals.
- Make capacity legible: queue depth, saturation, spend, per-workspace usage.
- Work with backend and psych-stack engineers on job orchestration from request to stored asset.
WHAT WE ARE LOOKING FOR
Mid-level platform/MLOps engineer who has run GPU workloads in production and can talk in dollars and milliseconds. Python plus infrastructure tooling; containers and IaC. US-based, remote or hybrid.
NICE TO HAVE
Diffusion/LLM serving at scale, spot/preemptible capacity, multi-region cost control.
HOW TO APPLY
There is nothing to sign up for and nothing to pay. Spend a little time on our website, (link removed), then tell us what you make of it - what looks strong, what looks thin, and what you would do differently in this role. Send your resume and those thoughts to .
Pixasocial LLC, 1209 Mountain Road PL NE Ste N, Albuquerque, NM 87110, USA. Equal-opportunity employer.
