AI Research Engineer

Research Engineer | San Francisco | Full-TimeBrief OverviewApplied AI lab building world models for 3D game environments. Early-stage, well-backed, 3 weeks from beta launch.What You'll OwnYou'll take large ownership of the core generation and simulation stack—the intelligence layer that turns ideas into playable 3D environments without any technical skill required.Your focus: distributed multi-agent orchestration for coordinated world synthesis (terrain, assets, behaviors), data-efficient training pipelines, real-time inference optimization, and agentic codegen tools that enable studio-quality environment creation.You'll architect systems that let creators go from concept to playable world in hours, not months. You'll define new benchmarks for retention, fun, and interactive intelligence using behavioral data across thousands of live environments. Your code ships to millions within weeks.Your ProfileStrong software fundamentals. You move fluidly between production systems and research code. You've shipped at scale.Multimodal & agentic systems. You've built or trained large models (diffusion, vision-language, RL agents). Hands-on with Claude, OpenAI, or custom implementations.Distributed training & scale. You've worked large-scale vision/3D pipelines, data curation, or distributed training. You know where things break and how to fix them.Preferred: Generative vision, world models, RL systems, robotics, game engines, or multi-agent coordination.The RealitySchedule: 996 for the first 4 months. Fast-moving team, tight deadlines.Location: San Francisco, on-site. Relocation support + free housing, food, transportation provided.Team: 12-person applied AI team. Flat structure, direct access to founders.Impact: Your work ships to millions of users within weeks.CompensationBase: $200K–$350K (flexible for exceptional candidates)Equity: $165.75K–$353.25K + benefitsAll-in TC: Up to $1M+ for top talent, heavily weighted to equityOther: Visa sponsorship, relocation support, free housing/food/transportationWhy This MattersYou're not training another large model or optimizing a search ranking. You're building the systems that let millions of people create, play, and monetize worlds together. Your work ships in weeks, not quarters.How It WorksVibe check (20 min with founder) — Mission fit, communication, intensity readinessTechnical deep dive (90 min with tech lead) — Systems design, research depth, shipping experienceInternal panel (60–90 min with team leads) — Collaboration, vision, ambiguity toleranceTrial task (1–2 days) — Real problem from the stack; shows your thinking + culture fitOffer — If aligned, fast moveTotal timeline: 10–15 days from first call to offer.