we generate in ~2 seconds. we want faster. speed is the product — it's what makes creation feel like play. you own the full ML pipeline. Josh is deep in infra and will be right there with you.what you'll do- own inference end to end — speed, quality, cost- run diffusion models in production and make them faster- rollouts, A/B testing, fallbacks, monitoring- evaluate new models fast, ship the ones that winstack- Python / PyTorch- diffusion models + low-latency LLMs- AWS GPU infra, inference optimizationwhat we want- you care about what users feel, not benchmark scores- you own production, not just experiments- you think in systems — latency, cost, quality, scale