PhD Machine Learning & Vision Scientist

Senior Machine Learning & Vision Scientist

About AIM Robots

AIM Robots build AI-native industrial robots that learn from video, understand complex manufacturing environments, and execute skilled physical tasks autonomously.

We are seeking exceptional PhD-level researchers—especially from world-class groups such as Yann LeCun’s NYU/FAIRlab, Stanford U, OpenAI, GoogleDeepMind—to define how robots see, learn, plan, and act in real production environments.

This is a hybrid research–engineering role with both scientific freedom and immediate real-world deployment impact.

Role Type

Research–Engineering Hybrid 

Full-time or Part-time 

Founding Scientist Track

What You Will Lead

You will architect AIM Robots’ next-generation perception, representation learning, real world knowledge abstration and world-modeling foundation.

You will design and deploy:

• Multi-modal models
• Video understanding systems
• World models for prediction and planning
• Robust policy-learning pipelines for autonomous robot behavior in real factories

You will work closely with the founding team to shape both the research roadmap and the production AI stack.

Core FocusAreas

1. Perception & Video Understanding
   • Design 2D/3D perception pipelines for complex, cluttered manufacturing environments
   • Fuse information from multiple cameras, RGB-D sensors, and other modalities
   • Develop video models that understand temporal structure and operator actions
   • Convert raw human demonstrations into structured, machine-interpretable representations
2. Representation Learning & Self-Supervision
   • Build scalable self-supervised learning (SSL) pipelines
     • (e.g., JEPA / I-JEPA,YOLO, DINO, SAM, MAE, MoCo)
   • Develop efficient representation backbones optimized for real-time performance and reliability
   • Explore structured intermedidate and other advanced representation-learning and understanding approaches that tightly link perception and controls
3. World Models & Autonomous Behavior
   • Develop multi-modal world models that combine perception, dynamics, and control
   • Build action-conditioned generative models (e.g., transformers, diffusion) for prediction and planning
   • Enable multi-step reasoning and planning for complex physical tasks
   • Drive generalization across workflows, products, and factories
4. Imitation Learning, RL & PolicyGeneration
   • Develop behaviorcloning, DAgger, and offline RL pipelines from demonstration data
   • Build diffusion-policy or related generative control models for fine manipulation and robust policies
   • Extract reusable, generalizable behaviors from human demonstration video
   • Collaborate with robotics engineers to deploy and iterate policies in production environments
5. Sim2Real, Synthetic Data & TrainingInfrastructure
   • Design synthetic data generation workflows for perception and policy learning
   • Use domain randomization and physics-based variation to improve robustness under distribution shifts
   • Integrate digitaltwins and simulation environments to accelerate large-scale training
   • Create tools and pipelines for training, evaluating, and monitoring multi-modal models at scale

Qualifications Required

• PhD (or equivalent research track) in Machine Learning, Computer Vision, Robotics, or a closely related field
• Expertise in self-supervised learning, representation learning, video modeling, or world models
• Strong fundamentals in 2D/3D perception and modern deep learning
• Experience training and scaling deep learning systems on real data
• Familiarity with robotics, embodied AI, or ML systems that operate in the physical world

Nice to Have

• Experience with Deep Learning models, VLM/VLA or world-model research
• Strong engineering skills in PyTorch, C++, CUDA, or Triton
• Background in top-tier research labs (e.g., NYU, FAIR, DeepMind,Google Brain)
• Experience working with large-scale multi-modal datasets or real robot data

Compensation & Track

We offer a competitive salary and meaningful equity. The Founding Scientist track includes:

• Enhanced equity grants
• Optional participation in revenue tied to developed AI modules and deployed systems
   

Why AIM Robotics

• Your research goes directly into real U.S. factories and production systems
• Access to large-scale multi-modal data: video, depth, sensor streams, and more
• Opportunity to design the intelligence stack for next-generation industrial robots
• Blend of academic-style research freedom with fast, pragmatic engineering execution
• High-impact role influencing technical strategy, system architecture, and hiring