Hiring || Spatial Reasoning Evaluation Specialist (Video RL)

ROLE OVERVIEW & OBJECTIVE:

We are seeking specialized Spatial Reasoning Evaluation Specialists to join our flagship Video Reinforcement Learning programme. This program focuses directly on physical-AI simulations, world models, autonomous agent trajectories, and avatar motion synthesis. In this role, you will perform rigorous step-wise trajectory reviews, auditing temporal ordering, physical plausibility, action anticipation, and segmentation mask fidelity. You will establish the human ground truth required to unlock next-generation spatio-temporal AI models. . 2.

KEY RESPONSIBILITIES & CORE WORKFLOWS

• Step-Wise Trajectory Review: Conduct granular, frame-by-frame evaluations of agentic video rollouts, classifying physics anomalies, causal inconsistencies, and execution errors.
• Temporal Logic & Sequence Verification: Audit complex multi-event chronological sequences, ensuring correct action-effect causality and forward-backward temporal logic.
• Action Anticipation & Goal Reasoning: Evaluate whether autonomous agent behaviors align with specified multi-step goals, flagging path deviations or suboptimal sub-goals.
• Video Segmentation & Rendering QA: Verify spatial boundary boxes, optical flow fields, pixel-level semantic segmentation masks, and avatar skeletal motion rendering.
• Deep Rationale & Error Taxonomy: Author highly structured C1-level failure-mode analyses categorizing visual hallucinations, unnatural physics, and clipping artifacts.
• Golden Dataset Construction: Build hardened ground-truth validation sets and edge-case scenario libraries to boost programmatic verifier accuracy.

CANDIDATE PROFILE & QUALIFICATIONS

Mandatory Requirements

• Educational & Technical Background: Bachelor's or Master's degree in a STEM discipline (Computer Science, Engineering, Physics, Data Science, Mathematics) or equivalent technical experience.
• Domain Exposure: Familiarity with video analysis, computer vision, reinforcement learning (RL) paradigms, or autonomous robotics/simulation environments.
• Cognitive Stamina: Exceptional spatial-temporal reasoning, strong visual attention to detail, and the focus required for deep task evaluation cycles.
• Language Mastery: C1 written English proficiency for drafting detailed, highly analytical technical evaluations and logic proofs.

Preferred Qualifications

• Scripting & Automation: Basic scripting capability in Python for data verification and log parsing.
• ML Concepts: Foundational understanding of reward models, verifier networks, diffusion models, or 3D Gaussian splatting.
• Annotation Platforms: Experience using advanced video annotation tooling (CVAT, Labelbox, Roboflow).
• Language Mastery: C1 written English proficiency for drafting detailed, highly analytical technical evaluations and logic proofs.