As the Technical Program Manager for our World Model team, you will be the operational backbone of IFM's most ambitious research frontier: PAN (Physical, Agentic, Nested). While most world models focus on generating static, visual 3D scenes, PAN is focused on interactive reasoning and long-horizon simulation with environments you can actually act in.
We are seeking a highly technical operator who is deeply passionate about the future of interactive AI and has been an avid user and follower of early generation models (e.g., Sora and similar platforms). You will impose the rigor required to scale PAN-v2, translating cutting-edge Generative Latent Prediction (GLP) research into clear, actionable training milestones. You will dive directly into the engine room by managing compute allocation, unblocking data pipelines, scaling our evaluation frameworks, and ensuring our researchers have the operational support they need to maintain high-velocity delivery.
Key Responsibilities
Operational Rigor & Execution: Own the tactical execution of the World Model training cycles. Manage the day-to-day milestones for PAN-v2, tracking everything from dataset curation to distributed training runs.
Clear Technical Bottlenecks: Actively manage and scale our most critical operational tracks, specifically focusing on expanding our Data Pipelines and standing up comprehensive Evaluation frameworks.
Technical Translation: Bridge the gap between Researchers, Data Scientists, and Infrastructure Engineers. You must understand the nuances of managing jobs, model checkpoints, and compute quotas to quickly resolve issues and prevent downtime.
Force-Rank Priorities: Ruthlessly prioritize engineering and research tasks to keep the team on the critical path for scheduled model releases. Convert flexible research timelines into committed delivery targets.
Cross-Functional Synchronization: Ensure that the World Model outputs seamlessly integrate with our agent, reasoning, and deployment teams.
Academic Qualifications
Professional Experience- Minimum
5+ years of experience in Technical Program Management, Engineering Management, Product Management, deep R&D environments, or equivalent.
Technical Operator: You possess deep familiarity with the ML training lifecycle. You understand the difference between pre-training and inference, know what a checkpoint is, and have experience supporting large-scale distributed training runs.
Proven "Shipper": A portfolio of complex technical projects you have successfully managed from conception to release.
Operational Fluency: Strong working knowledge of project estimation, risk management, and issue tracking for highly complex, ambiguous technical systems.
Professional Experience- Preferred
World Model Passion: You are a genuine enthusiast and avid user of world models, video generation tools (e.g., Sora), or simulation engines. You possess a strong baseline understanding of the current landscape of interactive reasoning models.
Product Management Interest: While this is an operational role, experience in or a strong desire to contribute to Product Management is a major plus. You are eager to help shape the future applications, product vision, and downstream use cases for our world models.
Foundation Model Expertise: Experience working alongside researchers utilizing diffusion models, autoregressive LLMs, or reinforcement learning frameworks.
