Research Engineer ($200k - $350k + equity) at Roam

This is a job that Standout, the AI Career agent, is recruiting for on behalf of one of our customers.Standout will pick the best talents from the ones he represents.The next step is to speak to Standout.Who We're Hiring For: Roam is an Applied AI lab building World Models for 3D environments, starting with gaming: letting anyone turn an idea into a playable game without technical skills.Role Snapshot: Research Engineer.You'll build and optimize the generation pipelines, training methods, and inference systems that produce interactive 3D worlds in real time.Compensation: $200k - $350k + equityFree housing, food, and transportation included.Visa sponsorship and relocation support available.What This Role Is About:You'll work across Roam's core AI stack: distributed multi-agent orchestration for world synthesis, RL pipelines for adaptive game generation, and agentic codegen and diffusion tooling for 3D environments. The team is 12 people, built from ex-Moonshot researchers, second-time founders, and engineers who shipped 100+ games at studios like Voodoo and Supersonic. Your output goes directly into products that 1,000+ creators with 250M+ combined reach will use to generate and publish games.Location: San Francisco, onsite.What Makes This Opportunity Exceptional:Roam has already locked in 1,000+ gaming and livestream creators (100K to 10M followers each) as launch partners, giving your work a distribution channel of 250M+ fans from day one.The team acquired a consumer product studio with 11.5M+ downloads, so you're building on real product traction, not starting from a research paper.Backed by investors behind ByteDance and Anduril, Together AI, and Polymarket.What You'll Be Responsible For:Build and scale distributed training, data curation, and inference pipelines for multimodal 3D world generation, from terrain synthesis to asset population and behavior trees.Design and run RL pipelines for open-ended game generation with hierarchical action spaces and dynamic difficulty adjustment, defining new benchmarks for retention and interactive intelligence.Own agent orchestration and diffusion-based tooling that turns creator inputs into studio-quality simulated environments ready for real-time play.Who Will Thrive in This Role:You switch comfortably between research and production code, and you've built or trained large multimodal or agentic systems with real scale constraints (distributed training, large vision or 3D pipelines).You want outsized ownership early. With a 12-person team, you'll own major parts of the stack, which means fewer guardrails and more ambiguity than a structured org.You are genuinely fine with a 996 schedule in office. This is not a figure of speech. The pace is intense and sustained, and the team is upfront about it.How to Apply:Apply in 2 minutes via Standout, your AI career agentClick "Apply with Standout"Have a chat with the agentIf its a fit get fast tracked directly to the hiring teamIf not get matched with other opportunities in Standout network